Label-Synchronous Speech-to-Text Alignment for ASR Using Forward and Backward Transformers.

Yusuke Kida Tatsuya Komatsu Masahito Togami

Published in: CoRR (2021)

Keyphrases

forward and backward
automatic speech recognition
multi label
speech recognition
greedy search
dynamic programming
noisy environments
objective function
lower bound
evolutionary algorithm
computational efficiency