The dataset requires cleaning.

#1
by dzja112 - opened

Random listening revealed numerous artifacts within the dataset: words spoken in square brackets instead of non-verbal vocalizations; noise and etc artifacts instead of audible speech. Training a neural network on such material can lead to pronunciation errors and artifacts

Sign up or log in to comment