README.md

---
language:
- gl
license: apache-2.0
base_model: openai/whisper-large
tags:
- whisper-event
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_13_0
metrics:
- wer
model-index:
- name: Whisper Large Galician
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: mozilla-foundation/common_voice_13_0 gl
      type: mozilla-foundation/common_voice_13_0
      config: gl
      split: test
      args: gl
    metrics:
    - name: Wer
      type: wer
      value: 6.939845474613686
---

# Whisper Large Galician

## Model summary

**Whisper Large Galician** is an automatic speech recognition (ASR) model for **Galician (gl)** speech. It is fine-tuned from [openai/whisper-large] on the **Galician portion of Mozilla Common Voice 13.0**, achieving a **Word Error Rate (WER) of 6.94%** on the Common Voice evaluation split.

This model provides high-accuracy transcription for large-scale Galician ASR applications.

---

## Model description

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

The large model leverages Whisper’s multilingual pretraining and is fine-tuned on Galician speech data to deliver high-quality transcription suitable for research, media, and accessibility applications.

---

## Intended use

### Primary use cases

* High-accuracy transcription of Galician audio recordings
* Offline or batch ASR pipelines
* Research and development in Galician ASR
* Media, educational, and archival transcription tasks

### Intended users

* Researchers working on Galician or low-resource ASR
* Developers building Galician speech applications
* Academic or institutional users

### Out-of-scope use

* Real-time or low-latency ASR without 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
* Transcription errors may 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 (Galician 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 (Galician evaluation split)
* **Metric:** Word Error Rate (WER)

---

## Evaluation results

| Metric     | Value      |
| ---------- | ---------- |
| WER (eval) | **6.94%** |

This reflects the expected performance of a large Whisper model fine-tuned for Galician.

---

## 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: 20,000
* Train batch size: 32
* Evaluation batch size: 16
* Gradient accumulation steps: 2
* Total train batch size: 64
* Seed: 42

### Training results (summary)

| Training Loss | Epoch | Step  | Validation Loss | WER    |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| 0.0126        | 4.01  | 1000  | 0.2128          | 8.3558 |
| 0.0032        | 9.01  | 2000  | 0.2262          | 6.9416 |
| 0.0022        | 14.01 | 3000  | 0.2528          | 7.1123 |
| 0.0025        | 19.01 | 4000  | 0.2643          | 7.3641 |
| 0.0015        | 24.01 | 5000  | 0.2596          | 7.3365 |
| 0.0014        | 29.01 | 6000  | 0.2723          | 7.6366 |
| 0.0008        | 34.01 | 7000  | 0.2778          | 7.6090 |
| 0.0003        | 39.01 | 8000  | 0.2880          | 7.2261 |
| 0.0004        | 44.01 | 9000  | 0.2920          | 7.6745 |
| 0.0001        | 49.01 | 10000 | 0.2854          | 7.4089 |
| 0.0           | 54.01 | 11000 | 0.3027          | 7.4365 |
| 0.0           | 59.01 | 12000 | 0.3159          | 7.4055 |
| 0.0           | 64.01 | 13000 | 0.3242          | 7.3693 |
| 0.0           | 69.01 | 14000 | 0.3312          | 7.3072 |
| 0.0           | 74.01 | 15000 | 0.3379          | 7.0226 |
| 0.0           | 79.01 | 16000 | 0.3442          | 7.0019 |
| 0.0           | 84.01 | 17000 | 0.3500          | 6.9933 |
| 0.0           | 89.01 | 18000 | 0.3550          | 6.9605 |
| 0.0           | 94.01 | 19000 | 0.3589          | 6.9467 |
| 0.0           | 99.01 | 20000 | 0.3605          | 6.9398 |

---

## 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-large-gl"  # 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](https://arxiv.org/abs/2503.23542)
for more details.

---

## License

This model is available under the
[Apache-2.0 License](https://www.apache.org/licenses/LICENSE-2.0).
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.