L. Ferrer, D. Castan, M. McLaren, and A. Lawson, "A Hierarchical Model for Spoken Language Recognition", arXiv:2201.01364, 2021
Table_of_stats (in pdf and xlxs formats) shows the number of samples in each train/development/evaluation dataset used in the paper for each language. For dev and eval datasets, we include separately the counts for the 8 - and 32-second chunks. For the training datasets, the count corresponds to only the original raw signals, before augmentation and chunking. In red are the lines for languages for which no training data is available (ie, out of set languages).