The benchmark is done on 6 french datasets:
- Common Voice
- Multilingual LibriSpeech
- VoxPopuli
- SUMM-RE
- TEDx
- YouTube
The following table gives the result of the WER on several french datasets:
The code to re create the results and the figure are available in this folder.