Request for Information on Dataset Used in Quality Assessment
Hello! I really appreciate the impressive work you've done on this project. Could you please tell me which dataset you used to calculate the quality metrics (WER, CER)?
Hello @Lightmourne , I hope this message finds you well.
The model was evaluated on oza75/bambara-asr, using the clean_combined config and the test split, which has a total duration of over 6 hours. During evaluation, both the inputs and outputs were normalized.
The dataset is primarily based on the jeli-asr dataset.
Franchement grand merci à toi Sudoping01 car je l'ai testé impeccable contrairement à mon 1er model basé sur small kalilouisangare/whisper-small-bambara-v2-kis avec un taux d'erreur de 44% qui est très loin du compte de cette version large 25% même avec la conversion ct2 pour optimisation reste significatif.
@kalilouisangare glad to hear the model performs well for your use case! Thanks for sharing the feedback