|
--- |
|
base_model: openai/whisper-small |
|
datasets: |
|
- mozilla-foundation/common_voice_17_0 |
|
language: gl |
|
library_name: transformers |
|
license: apache-2.0 |
|
model-index: |
|
- name: Finetuned openai/whisper-small on Galician |
|
results: |
|
- task: |
|
type: automatic-speech-recognition |
|
name: Speech-to-Text |
|
dataset: |
|
name: Common Voice (Galician) |
|
type: common_voice |
|
metrics: |
|
- type: wer |
|
value: 13.681 |
|
--- |
|
|
|
# Finetuned openai/whisper-small on 35141 Galician training audio samples from mozilla-foundation/common_voice_17_0. |
|
|
|
This model was created from the Mozilla.ai Blueprint: |
|
[speech-to-text-finetune](https://github.com/mozilla-ai/speech-to-text-finetune). |
|
|
|
## Example |
|
|
|
Speech input: |
|
|
|
<audio controls><source src="https://huggingface.co/mozilla-ai/whisper-small-gl/resolve/main/gl-example.wav" type="audio/wav"></audio> |
|
|
|
Text output: |
|
|
|
| Ground Truth | [openai/whisper-small](https://huggingface.co/openai/whisper-small) | [mozilla-ai/whisper-small-gl](https://huggingface.co/mozilla-ai/whisper-small-gl) *| |
|
| -------------| -------------| ------------------- | |
|
| O Comit茅 Econ贸mico e Social Europeo deu luz verde esta terza feira ao uso de galego, euskera e catal谩n nas s煤as sesi贸ns plenarias, segundo informou o Ministerio de Asuntos Exteriores nun comunicado no que se felicitou da decisi贸n. | O Comit茅 Econ贸mico Social Europeo de Uluz Verde est谩 terza feira a Ousse de Gallego e Uskera e Catalan a s煤as asesi贸ns planarias, segundo informou o Ministerio de Asuntos Exteriores nun comunicado no que se felicitou da decisi贸n. | O Comit茅 Econ贸mico Social Europeo deu luz verde esta terza feira ao uso de galego e usquera e catal谩n nas s煤as sesi贸ns planarias, segundo informou o Ministerio de Asuntos Exteriores nun comunicado no que se felicitou da decisi贸n. | |
|
|
|
|
|
## Evaluation results on 9990 audio samples of Galician: |
|
|
|
### Baseline model (before finetuning) on Galician |
|
- Word Error Rate: 40.812 |
|
- Loss: 1.506 |
|
|
|
### Finetuned model (after finetuning) on Galician |
|
- Word Error Rate: 13.681 |
|
- Loss: 0.21 |
|
|