Goobo Labs
Research
OverviewPublicationsBenchmarks (SomBench)Open SourceRoadmap
DatasetsModels
Learn
ProgramsLearning PathsCurriculum
BlogCommunity
Sign inExplore research
Goobo Labs

An open research lab building the foundations of Somali AI.

Research

  • Overview
  • Publications
  • Benchmarks
  • Roadmap

Datasets

  • SomNLP Corpus
  • SomNLP STT
  • Wikipedia Corpus
  • Citation Guide

Programs

  • Data Science & ML Bootcamp
  • Python for Everyone
  • NLP (coming soon)
  • Curriculum

Resources

  • Docs
  • Models & Tools
  • Open Source
  • Blog
  • Soplang

Contact

  • info@goobolabs.so
  • Mogadishu, Somalia

Company

  • About
  • Careers
  • Impact
  • Press Kit
  • Verify Certificate
  • Contact

Community

  • GitHub
  • Hugging Face
  • X
  • LinkedIn

© 2026 Goobo Labs. Open research for Somali AI.

Privacy PolicyTerms of Service
HomeBlogModel
ModelMay 27, 2026 · 6 min read

Releasing whisper-som-small: 11.4% WER on Somali radio speech

Our first public ASR checkpoint — fine-tuned from Whisper-small on 320 hours of aligned Somali audio. Details on training setup, data splits, and how to use it in two lines of Python.

Omar Tood

Omar Tood

Data Scientist

Today we're releasing whisper-som-small, our first public Somali ASR checkpoint. Fine-tuned from OpenAI's Whisper-small on 320 hours of aligned Somali audio, it achieves 11.4% WER on held-out radio speech — a 38% relative improvement over the zero-shot multilingual baseline.

Background

Whisper is a strong multilingual ASR model, but Somali is severely underrepresented in its training data. The zero-shot Whisper-small baseline on our radio test set gives 18.4% WER — usable in some contexts, but not reliable enough for production transcription or downstream NLP tasks that depend on accurate text.

Fine-tuning on in-domain data is the obvious fix. The challenge is data: high-quality aligned speech is expensive to produce. Our SomNLP-STT-Corpus, released last year, gave us the raw material to do this properly.

The dataset

SomNLP-STT-Corpus contains 320 hours of Somali audio across four sources: radio broadcasts (Radio Mogadishu, HornAfrik), government speeches, storytelling recordings, and community-contributed reading passages.

320h

Total audio

280h

Training split

20h

Validation

20h

Test split

Word-level alignment was generated using Montreal Forced Aligner with our custom Somali acoustic model, then manually corrected on a 10% random sample. Alignment accuracy on the corrected sample was 96%.

Training setup

We fine-tuned Whisper-small (244M parameters) for 8,000 steps on a single A100 40GB GPU, using Hugging Face transformers with Seq2SeqTrainer, a learning rate of 1e-5, linear warmup over 500 steps, and gradient accumulation over 4 steps for an effective batch size of 16.

We froze the encoder for the first 2,000 steps to preserve general acoustic representations before allowing end-to-end fine-tuning. This reduced early divergence and improved final WER by ~1.2 points over full fine-tuning from step 0.

Results

On the held-out 20-hour radio test set, whisper-som-small achieves 11.4% WER — down from 18.4% zero-shot. The absolute improvement is largest on broadcast news (−8.2%) and smallest on conversational speech (−4.1%), where our training data was thinner.

Usage

python
import whisper

# Load the fine-tuned model
model = whisper.load_model("goobolabs/whisper-som-small")

# Transcribe audio
result = model.transcribe("somali_audio.wav", language="so")
print(result["text"])
# 'Beeraha Soomaaliya waxay u baahan yihiin tignoolajiyad cusub.'

# With word-level timestamps
result = model.transcribe("audio.wav", language="so", word_timestamps=True)
for seg in result["segments"]:
    print(f"{seg['start']:.1f}s – {seg['end']:.1f}s: {seg['text']}")
ASRSpeechModelWhisper
Omar Tood

Omar Tood

Data Scientist

Writing about Somali language technology, open data, and AI from the lab in Mogadishu.

All postsGet in touch

On this page

  • Background
  • The dataset
  • Training setup
  • Results
  • Usage

Share

Back to Blog
More from the lab

Related posts

Research6 min read

Introducing the Somali Language Standard (SLS) – Phase 1 Complete

We've formally published the first two standards of the Somali Language Standard (SLS) — an open, machine-readable, CI/CD-validated framework for the Somali language, starting with the alphabet and the standards process itself.

Jul 9, 2026Read
Research6 min read

Soplang v2.0.0: a compiler for the first Somali programming language

Our team shipped a Rust-based compiler for Soplang — Cranelift JIT and ahead-of-time builds, with syntax written entirely in Somali keywords.

Jun 30, 2026Read
Dataset5 min read

SomNLP-Corpus v2: 887M+ tokens from six open sources

Our largest Somali text release yet: 1.77M clean documents, 591M words, and ~887M tokens from HPLT, CC100, mC4, OPUS, MADLAD, and MT560 — filtered through a six-stage pipeline.

Jun 9, 2026Read
Newsletter

Stay in the loop

Dataset releases, new bootcamp cohorts, research papers, and lab updates — straight to your inbox. No spam, unsubscribe anytime.