Accents mismatching is a critical problem for end-to-end ASR. This paper aims to address this problem by building an accent-robust RNN-T system with domain adversarial training (DAT). We unveil the magic behind DAT and provide, for the first time, a theoretical guarantee that DAT learns accent-invariant representations. We also prove that performing the gradient reversal in DAT is equivalent to minimizing the Jensen-Shannon divergence between domain output distributions. Motivated by the proof of equivalence, we introduce reDAT, a novel technique based on DAT, which relabels data using either unsupervised clustering or soft labels. Experiments on 23K hours of multi-accent data show that DAT achieves competitive results over accent-specific baselines on both native and non-native English accents but up to 13% relative WER reduction on unseen accents; our reDAT yields further improvements over DAT by 3% and 8% relatively on non-native accents of American and British English.
@article{arxiv.2012.07353,
title = {REDAT: Accent-Invariant Representation for End-to-End ASR by Domain Adversarial Training with Relabeling},
author = {Hu Hu and Xuesong Yang and Zeynab Raeesy and Jinxi Guo and Gokce Keskin and Harish Arsikere and Ariya Rastrow and Andreas Stolcke and Roland Maas},
journal= {arXiv preprint arXiv:2012.07353},
year = {2021}
}
Comments
accepted in ICASSP 2021; final camera-ready version