Abstract: Albeit great performance of Transformer-based speech self-supervised learning (SSL) models, their large parameter size and computational cost make them unfavorable to utilize. In this study, ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback