1 paper
Roger Hsiao, Liuhui Deng, Erik McDermott +2
We propose a novel approach to optimizing a byte-level representation for end-to-end automatic speech recognition (ASR). Byte-level representation is often used by large scale mult…