HiTZ/whisper-medium-eu

Whisper Medium Basque

Model summary

Whisper Medium Basque is an automatic speech recognition (ASR) model for Basque (eu) speech. It is fine-tuned from [openai/whisper-medium] on the Basque portion of Mozilla Common Voice 13.0, achieving a Word Error Rate (WER) of 14.12% on the Common Voice evaluation split.

This model offers a balance between transcription accuracy and computational requirements, providing significantly improved ASR performance over smaller Whisper variants while remaining practical for offline or batch processing.

---

Model description

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

This medium-sized model leverages Whisper’s multilingual pretraining and is fine-tuned on Basque speech data, delivering higher transcription quality for a low-resource language while remaining manageable for typical GPU or CPU environments.

---

Intended use

Primary use cases

• High-quality transcription of Basque audio recordings
• Offline or batch ASR pipelines
• Research and development in Basque ASR
• Media, educational, and archival transcription tasks

Intended users

• Researchers working on Basque or low-resource ASR
• Developers building Basque speech applications
• Academic and institutional users

Out-of-scope use

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

---

Limitations and known issues

• Performance may degrade on:
  • Noisy or low-quality recordings
  • Conversational or spontaneous speech
  • Accents underrepresented in Common Voice
• While highly accurate for a medium-sized model, errors can still occur under challenging acoustic conditions
• Dataset biases from Common Voice 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)	14.12%

These results indicate strong transcription performance for a medium-sized Whisper model fine-tuned for Basque.

---

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: 10,000
• Train batch size: 64
• Evaluation batch size: 32
• Seed: 42

Training results (summary)

Training Loss	Epoch	Step	Validation Loss	WER
0.0206	4.02	1000	0.2998	16.9995
0.0036	9.01	2000	0.3235	15.5211
0.0018	14.01	3000	0.3454	14.9905
0.0013	19.01	4000	0.3538	14.9439
0.0013	24.01	5000	0.3587	14.8568
0.0002	29.0	6000	0.3799	14.4153
0.0001	33.02	7000	0.3937	14.2067
0.0001	38.02	8000	0.4050	14.1946
0.0001	43.01	9000	0.4119	14.1196
0.0001	48.01	10000	0.4150	14.1358

---

Framework versions

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

---

How to use

from transformers import pipeline

hf_model = "HiTZ/whisper-medium-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:

@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.