基於注意力(Attention)機制的端到端系統,又被稱為LAS端到端構架。
[6] W. Chan, N. Jaitly, Q. Le, O. Vinyals. Listen, Attend and Spell: A Neural Network for Large Vocabulary Conversational Speech Recognition. ICASSP 2016.
來自 <https://mp.weixin.qq.com/s/c64XucML13OwI26_UE9xDQ>
為了更好地進行LAS模型的訓練。可以使用以下技術:
- Schedule Sampling
- Label Smoothing
- Multi-head Attention