Whisper Small Basque is an automatic speech recognition (ASR) model for Basque (eu) speech. It is fine-tuned from [openai/whisper-small] on the Basque portion of Mozilla Common Voice 13.0, achieving a Word Error Rate (WER) of 18.42% on the Common Voice evaluation split.

The model provides a balance between transcription accuracy and computational efficiency, offering substantially improved performance over tiny models while remaining suitable for offline and batch ASR use.

Model description

Architecture: Transformer-based encoder–decoder (Whisper)
Base model: openai/whisper-small
Language: Basque (eu)
Task: Automatic Speech Recognition (ASR)
Output: Text transcription in Basque
Decoding: Autoregressive sequence-to-sequence decoding

This model builds on Whisper’s multilingual pretraining and is further fine-tuned on Basque speech data to improve recognition quality for a low-resource language while maintaining moderate computational requirements.

Intended use

Primary use cases

Transcription of Basque audio recordings
Offline or batch ASR pipelines
Research on Basque and low-resource speech recognition
Media, educational, and archival transcription tasks

Intended users

Researchers working on Basque ASR
Developers building Basque speech applications
Academic and institutional users

Out-of-scope use

Real-time or low-latency ASR without further optimization
Speech translation tasks
Safety-critical or high-risk applications without additional validation

Limitations and known issues

Performance may degrade on:
- Highly noisy or low-quality recordings
- Conversational or spontaneous speech
- Accents underrepresented in Common Voice
While significantly more accurate than tiny models, it may still produce errors in challenging acoustic conditions
Biases present in the Common Voice dataset may be reflected in outputs

Users are encouraged to evaluate the model on their own data before deployment.

Training and evaluation data

Training data

Dataset: Mozilla Common Voice 13.0 (Basque subset)
Data type: Crowd-sourced, read speech
Preprocessing:
- Audio resampled to 16 kHz
- Text normalized using Whisper tokenizer
- Filtering of invalid or problematic samples

Evaluation data

Dataset: Mozilla Common Voice 13.0 (Basque evaluation split)
Metric: Word Error Rate (WER)

Evaluation results

Metric	Value
WER (eval)	18.42%

These results demonstrate a strong improvement over smaller Whisper variants for Basque ASR.

Training procedure

Training hyperparameters

Learning rate: 1e-5
Optimizer: Adam (β1=0.9, β2=0.999, ε=1e-8)
LR scheduler: Linear
Warmup steps: 500
Training steps: 5,000
Train batch size: 16
Evaluation batch size: 8
Seed: 42

Training results (summary)

Training Loss	Epoch	Step	Validation Loss	WER
0.2826	1.04	1000	0.3472	24.9342
0.0872	2.07	2000	0.3012	20.2661
0.0275	3.11	3000	0.3085	19.3021
0.0086	4.14	4000	0.3297	18.7513
0.0051	6.01	5000	0.3390	18.4171

Framework versions

Transformers 4.33.0.dev0
PyTorch 2.0.1+cu117
Datasets 2.14.4
Tokenizers 0.13.3

How to use

python

from transformers import pipeline

hf_model = "HiTZ/whisper-small-eu"  # replace with actual repo ID
device = 0  # set to -1 for CPU

pipe = pipeline(
    task="automatic-speech-recognition",
    model=hf_model,
    device=device
)

result = pipe("audio.wav")
print(result["text"])

Ethical considerations and risks

This model transcribes speech and may process personal data.
Users should ensure compliance with applicable data protection laws (e.g., GDPR).
The model should not be used for surveillance or non-consensual audio processing.

Citation

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

bibtex

@misc{dezuazo2025whisperlmimprovingasrmodels,
  title={Whisper-LM: Improving ASR Models with Language Models for Low-Resource Languages},
  author={Xabier de Zuazo and Eva Navas and Ibon Saratxaga and Inma Hernáez Rioja},
  year={2025},
  eprint={2503.23542},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}

Please, check the related paper preprint in arXiv:2503.23542 for more details.

License

This model is available under the Apache-2.0 License. You are free to use, modify, and distribute this model as long as you credit the original creators.

Contact and attribution

Fine-tuning and evaluation: HiTZ/Aholab - Basque Center for Language Technology
Base model: OpenAI Whisper
Dataset: Mozilla Common Voice

For questions or issues, please open an issue in the model repository.

Funding

This project with reference 2022/TL22/00215335 has been parcially funded by the Ministerio de Transformación Digital and by the Plan de Recuperación, Transformación y Resiliencia – Funded by the European Union – NextGenerationEU ILENIA and by the project IkerGaitu funded by the Basque Government. This model was trained at Hyperion, one of the high-performance computing (HPC) systems hosted by the DIPC Supercomputing Center.