https://arxiv.org/abs/2005.04862
Listen Attentively, and Spell Once: Whole Sentence Generation via a Non-Autoregressive Architecture for Low-Latency Speech Recognition (Ye Bai, Jiangyan Yi, Jianhua Tao, Zhengkun Tian, Zhengqi Wen, Shuai Zhang)
| 2005.04862
#non-autoregressive #asr