Using Data Augmentations and VTLN to Reduce Bias in Dutch End-to-End Speech Recognition Systems

TL;DR

This study employs data augmentation and VTLN to mitigate bias in Dutch end-to-end speech recognition, achieving notable reductions in error rates and bias across diverse speaker groups, with cross-language benefits demonstrated.

Contribution

It introduces a combined approach of data augmentation and VTLN for bias reduction in end-to-end speech recognition systems, demonstrating effectiveness across multiple speaker demographics and languages.

Findings

Reduced average WER by 6.9% with combined techniques.

Bias across speaker groups decreased by 3.9%.

VTLN trained on Dutch improved Mandarin child speech recognition.

Abstract

Speech technology has improved greatly for norm speakers, i.e., adult native speakers of a language without speech impediments or strong accents. However, non-norm or diverse speaker groups show a distinct performance gap with norm speakers, which we refer to as bias. In this work, we aim to reduce bias against different age groups and non-native speakers of Dutch. For an end-to-end (E2E) ASR system, we use state-of-the-art speed perturbation and spectral augmentation as data augmentation techniques and explore Vocal Tract Length Normalization (VTLN) to normalise for spectral differences due to differences in anatomy. The combination of data augmentation and VTLN reduced the average WER and bias across various diverse speaker groups by 6.9% and 3.9%, respectively. The VTLN model trained on Dutch was also effective in improving performance of Mandarin Chinese child speech, thus, showing generalisability across languages