Fetching the paper…

TERA: Self-Supervised Learning of Transformer Encoder Representation for Speech · Around