somali-speech-data / README.md
Professor's picture
Add dataset card
e30acb5 verified
|
Raw
History Blame Contribute Delete
4.54 kB
metadata
language:
  - so
license: cc-by-4.0
multilinguality: monolingual
task_categories:
  - text-to-speech
  - automatic-speech-recognition
tags:
  - somali
  - speech
  - tts
  - asr
  - african-languages
  - low-resource
pretty_name: Somali Speech Data (Pooled)
size_categories:
  - 10K<n<100K

Somali Speech Data (Pooled)

A ~103.2-hour Somali speech corpus, drawn from a single source (Afrivoice) and filtered to only genuinely transcribed audio. Part of the AfroNet multi-language TTS data effort.

Source

DigitalUmuganda/Afrivoice (the general, pan-African Afrivoice release — not Afrivoice_Ethiopia, which we've separately ingested for 5 Ethiopian languages) — Somali portion: 22,627 clips, 103.2h, source dataset_id/source = afrivoice.

There is also a Somali dataset published by African Next Voices (Anv-ke/Somali, 502h) — a completely different organization (Kenya-based KenCorpus Consortium vs. Rwanda-based Digital Umuganda). We use Afrivoice's Somali here; the two don't appear to be the same underlying recordings (their total-hours figures don't match closely, unlike the WAXAL-overlap cases described below), so Anv-ke's Somali release remains a candidate for a future, separate addition if more Somali hours are needed.

Why not WAXAL? WAXAL (google/WaxalNLP) has no Somali config at all, so there's no overlap question here (contrast with Shona/Lingala/Fulani/Malagasy, this source's other languages, which we've skipped for exactly that reason — see the project README).

A note on what "transcribed" means here

Afrivoice pairs each audio clip with an image the speaker was prompted to describe; transcription is the sentence the speaker was recorded saying. Only a minority of recordings are transcribed (~19% of total duration for Somali — 103.2h out of ~536h total) — the rest was recorded but never transcribed. No auto-transcription was used to unlock the untranscribed majority; only clips with a real, human-provided transcript are included here.

All audio is standardized to 16 kHz mono FLAC (lossless), 1–30 second clips. Source audio is real WAV — unlike Afrivoice_Ethiopia, there's no WebM-mislabeling bug here, and it decodes directly via soundfile with no ffmpeg step needed.

Format

The dataset ships as WebDataset-style tar shards (shards/shard-00000.tar …, ~1 GB each, one {key}.flac file per clip) plus a single manifest (manifest.parquet / manifest.jsonl):

Column Description
key, shard which tar file + entry holds this clip's audio
text transcript (native script), from Afrivoice's transcription field
duration seconds
source always afrivoice
dataset_id always 0
split train / val (250 clips held out for evaluation)
gender speaker metadata where available
dbfs, clip_ratio, sil_ratio cheap DSP quality proxies: loudness, fraction of clipped samples, fraction of near-silent frames
has_disfluency always false — this source doesn't flag disfluencies

Usage

from huggingface_hub import hf_hub_download
import pandas as pd, tarfile, io, soundfile as sf

mp = hf_hub_download("Professor/somali-speech-data", "manifest.parquet", repo_type="dataset")
df = pd.read_parquet(mp)

row = df.iloc[0]
shard_path = hf_hub_download("Professor/somali-speech-data", f"shards/{row.shard}", repo_type="dataset")
with tarfile.open(shard_path) as tar:
    audio_bytes = tar.extractfile(f"{row.key}.flac").read()
arr, sr = sf.read(io.BytesIO(audio_bytes))

The tar shards are also directly readable by the webdataset library for streaming training pipelines.

Intended use & limitations

Built for Somali TTS/ASR research, in particular as finetuning data for a multilingual TTS model that doesn't natively support Somali. Speech is prompted by an image-description task, a narrower register than natural conversation. This is a research aggregation; usage should respect Afrivoice's own terms.

License

CC BY 4.0, per the upstream Afrivoice release.

Acknowledgments

Deep thanks to Digital Umuganda for the Afrivoice corpus.

This dataset was pooled by Victor Olufemi and LyngualLabs as part of the AfroNet multi-language TTS data effort.