1 paper
Hainan Xu, Zhehuai Chen, Fei Jia +1
This paper proposes Transducers with Pronunciation-aware Embeddings (PET). Unlike conventional Transducers where the decoder embeddings for different tokens are trained independent…