Xingchen Song

end-to-end speech recognition (2) language model (2) data augmentation (1) speech recognition (1) speech processing (1) masked language model (1) automatic speech recognition (1) acoustic model (1) self-attention network (1) transformer encoder (1) weighted finite state transducer (1) contextual biasing (1) frequency channel (1) unified two-pass (1) streaming acoustic encoder (1) bi-directional representation (1) zero-shot learning (1) unified two-pass framework (1) unsupervised pretraining (1)