MCV-spoken-language-recognition

mix_40000_records.tsv.gz - train/test split for spoken language recognition on Mozilla Common Voice dataset

audio_models.py - models used for spoken language recognition on Mozilla Common Voice dataset

tuplemax_loss.py - Pytorch implementation of Tuplemax loss from Wan, Li, et al. "Tuplemax loss for language identification." ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2019.

See my article about spoken language recognition on Medium:

1. Data selection and preprocessing
2. Choosing optimal model
3. Audio transformations