---
language:
- ca
license: apache-2.0
base_model: openai/whisper-medium
tags:
- whisper-event
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_13_0
metrics:
- wer
model-index:
- name: Whisper Medium Catalan
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: mozilla-foundation/common_voice_13_0 ca
      type: mozilla-foundation/common_voice_13_0
      config: ca
      split: test
      args: ca
    metrics:
    - name: Wer
      type: wer
      value: 5.995427264932838
---

# Whisper Medium Catalan

## Model summary

**Whisper Medium Catalan** is an automatic speech recognition (ASR) model for **Catalan (ca)** speech. It is fine-tuned from [openai/whisper-medium] on the **Catalan subset of Mozilla Common Voice 13.0**, achieving a **Word Error Rate (WER) of 5.995%** on the evaluation split.

This model balances transcription accuracy and speed, offering higher performance than small variants while remaining computationally efficient.

---

## Model description

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

Fine-tuned to improve transcription quality on Catalan audio.

---

## Intended use

### Primary use cases

* High-accuracy transcription of Catalan audio  
* Research and development in Catalan ASR  
* Media, educational, or accessibility applications  

### Out-of-scope use

* Real-time transcription without optimization  
* Speech translation  
* Safety-critical applications without further validation  

---

## Limitations and known issues

* Performance may degrade on:
  * Noisy or low-quality recordings  
  * Conversational or spontaneous speech  
  * Regional dialects not well represented in Common Voice  
* Occasional transcription errors on difficult audio  

---

## Training and evaluation data

* **Dataset:** Mozilla Common Voice 13.0 (Catalan 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 metric:** Word Error Rate (WER) on held-out evaluation set  

---

## Evaluation results

| Metric     | Value      |
| ---------- | ---------- |
| WER (eval) | **5.995%** |

---

## 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  
* Eval batch size: 32  
* Seed: 42  

### Training results (summary)

| Training Loss | Epoch | Step  | Validation Loss | WER    |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| 0.1158        | 1.05  | 1000  | 0.1846          | 8.3630 |
| 0.0184        | 3.05  | 2000  | 0.2017          | 8.0629 |
| 0.0522        | 5.04  | 3000  | 0.1940          | 8.1177 |
| 0.0595        | 7.04  | 4000  | 0.1742          | 7.4696 |
| 0.0179        | 9.04  | 5000  | 0.1899          | 7.3095 |
| 0.0646        | 11.04 | 6000  | 0.1555          | 6.3441 |
| 0.0825        | 13.03 | 7000  | 0.1810          | 6.4841 |
| 0.0309        | 15.03 | 8000  | 0.1464          | 6.3544 |
| 0.0695        | 17.03 | 9000  | 0.1434          | 5.9954 |
| 0.0186        | 19.03 | 10000 | 0.1706          | 6.1097 |

---

## 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-medium-ca"  # 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.