LatticeBART: Lattice-to-Lattice Pre-Training for Speech Recognition.

Lingfeng Dai, Lu Chen, Zhikai Zhou, Kai Yu

Browse the full ICASSP paper archive.