phoneticoai/phonetico-speech

Phonetico Speech

v2605 · Tigrinya · 14.7 hours · CC-BY-4.0

Phonetico Speech is a speech corpus for automatic speech recognition (ASR) in Ethiopian languages. Each language is available as a separate config. Load only what you need. v2605 contains 14.7 hours of transcribed Tigrinya audio.

This dataset is part of a long-term effort to build foundational speech technology for Ethiopian languages.

Dataset Summary

Data Fields

Each row corresponds to one audio segment.

Data Splits

Speaker-disjoint. Gender-stratified. Test split is frozen across versions.

Versioning Policy

Phonetico Speech uses sequential version numbers in YYMM format (e.g., v2605 = May 2026). Each version is a complete snapshot of the corpus, tagged in the repository. Version numbers encode the release date and are strictly sequential. A higher number is always a newer release.

Any release may add new languages, expand existing languages, or correct transcription errors. The changelog documents what changed in each version.

Every metadata row includes a version field indicating when that segment was introduced. Prior versions are accessible via git tags:

ds = load_dataset("phoneticoai/phonetico-speech", "tir", revision="v2605")

Languages

Each language is a separate config. Load only the language you need.

Ethical Considerations

Speaker Consent and Compensation

All audio in this release was collected under explicit informed consent, with speakers aware that their recordings would be used for speech technology development and released publicly under CC-BY-4.0. Contributors were compensated for their work.

Data Sovereignty

Phonetico operates on a consent-first data sovereignty model. Speakers retain the right to request removal of their CC-BY-4.0 contributions. Removal requests are honored in the next release, with affected segment IDs documented in the changelog.

Bias and Representation

At 14.7 hours, this dataset does not claim to represent the full diversity of Tigrinya speech.

This release is predominantly read speech. Conversational and spontaneous speech will be added in future versions.

Intended Use

This dataset is intended for:

• —Training and evaluating Tigrinya ASR systems
• —Cross-lingual speech and translation research
• —Linguistic research on Tigrinya
• —Benchmarking ASR methods

This dataset is not intended for speaker identification, surveillance, or any application that attempts to re-identify anonymized speakers.

How to Use

With Hugging Face Datasets

from datasets import load_dataset

# Load a specific language
ds = load_dataset("phoneticoai/phonetico-speech", "tir", split="train")

# Pin to a specific version
ds = load_dataset("phoneticoai/phonetico-speech", "tir", split="train", revision="v2605")

# Access a sample
sample = ds[0]
audio = sample["audio"]
transcript = sample["text"]

Citation

If you use this dataset in your research, please cite:

@misc{phonetico-speech,
  title     = {Phonetico Speech: A Speech Corpus for Ethiopian Language ASR},
  author    = {Feyissa, Solen},
  year      = {2026},
  url       = {https://huggingface.co/datasets/phoneticoai/phonetico-speech},
  note      = {v2605}
}

License

This dataset is released under the Creative Commons Attribution 4.0 International License.

You are free to share and adapt this dataset for any purpose, including commercial use, provided you give appropriate credit.

Contact

• —Organization: Phonetico
• —Website: https://phonetico.ai
• —Email: data@phonetico.ai
• —Issues: Hugging Face discussion tab

See CHANGELOG.md for version history.