Other people's Ethiopian language datasets
Speech, text and translation data published by other researchers and organisations. None of it is ours. It is gathered here so that anyone starting work on an Ethiopian language has one place to look instead of searching and hoping.
These are third-party datasets. Read this first.
None of this is our data, and we have not checked its quality. Every entry belongs to the researcher or organisation that published it. We have not listened to it, measured it, or verified that it contains what its authors say it contains. This is a directory to make things findable, nothing more. Judge each one yourself, and credit its authors, not us.
We host none of it. Every link goes to the publisher, and the licence shown is the one they declared, copied rather than inferred. Where a publisher declared no licence we say so, because that means you have no permission to use it until you ask them.
Links and licences were last confirmed by hand on 27 August 2026. Things move. If you find something out of date or missing, tell us and we will fix it.
1 of the 13 entries below declare no licence at all. They are marked, and you should treat them as all rights reserved.
Speech
5 entries
ALFFA Amharic
MITALFFA project, via hadamard-2 · 10K–100K rows
The classic academic Amharic read-speech corpus for ASR, converted to the Hugging Face datasets format.
Long-standing reference corpus; several published Amharic ASR models are trained on it, so it is a reasonable baseline to compare against.
Amharic Speech Dataset 110HRS
None declaredKYAGABA · 10K–100K rows
One of the larger community-contributed Amharic speech collections.
The name claims 110 hours; the dataset card declares no licence and no hour count. Without a licence you have no permission to use it, so contact the author before building anything on it.
Common Voice 17 (Amharic and Tigrinya)
CC0 1.0Mozilla, via fsicoli
Mozilla's crowd-sourced read-speech corpus, with per-clip age, gender, accent and up/down vote metadata.
This is an unofficial community conversion, not Mozilla's own upload. The per-clip metadata schema is close to ours and is a useful reference point.
Amharic ASR Dataset
Apache 2.0Ephrem · Under 1K rows
A small community-contributed Amharic ASR set.
Tagged under 1,000 rows, so treat it as a sample rather than a training corpus.
Amharic Speech (protocol demonstration)
CC BY-NC 4.0Silencio Network · Under 1K rows
A small speaker-metadata-rich set published to demonstrate a collection protocol rather than to train models.
Non-commercial licence, so it cannot be used to train a commercial model. Its card openly documents gender and regional imbalance, which is a good template for how any corpus should state its own limits.
Model
3 entries
SpeechBrain DVoice Amharic ASR
Apache 2.0SpeechBrain
A wav2vec2 + CTC Amharic ASR system trained on the ALFFA corpus.
A model rather than a dataset. Listed because it is one of the few ready-to-run Amharic ASR baselines.
Ethio-ASR
See individual modelsBadr al-Absi (badrex)
A family of wav2vec2-BERT ASR models covering five Ethiopian languages, with both per-language and multilingual variants.
The widest open ASR coverage of Ethiopian languages we are aware of. We benchmark against these.
Amharic TTS (SpeechT5)
MITAddisuSeteye
A SpeechT5 model fine-tuned for Amharic speech synthesis.
Text
3 entries
Amharic Sentences Corpus
Apache 2.0a3xrfgb · 1M–10M rows
A large Amharic sentence corpus extracted from public Telegram channel exports.
The extraction tooling is open source, so the corpus can be reproduced and extended.
Amharic News Text Classification
CC BY 4.0Israel Abebe Azime · 10K–100K rows
A labelled Amharic news corpus with topic categories (Azime & Mohammed, 2021).
Widely cited. Also mirrored on GitHub.
Amharic News Category Classification
CC BY 4.0rasyosef · 10K–100K rows
Amharic news articles labelled by category, also usable for summarisation.
Translation
1 entry
Directory
1 entry
Know one we are missing?
This list is short because it only contains things we have opened and confirmed by hand. Send us anything that belongs here, including your own work. We will list datasets that compete with ours: a researcher who cannot find data is a worse outcome than a researcher who picks someone else's.