Fetching the paper…

Speech-XLNet: Unsupervised Acoustic Model Pretraining For Self-Attention Networks · Around