Showing 1–1 of 1 results for author: Noguero, D S

Speech Robust Bench: A Robustness Benchmark For Speech Recognition

Authors: Muhammad A. Shah, David Solans Noguero, Mikko A. Heikkila, Nicolas Kourtellis

Abstract: As Automatic Speech Recognition (ASR) models become ever more pervasive, it is important to ensure that they make reliable predictions under corruptions present in the physical and digital world. We propose Speech Robust Bench (SRB), a comprehensive benchmark for evaluating the robustness of ASR models to diverse corruptions. SRB is composed of 69 input perturbations which are intended to simulate… ▽ More As Automatic Speech Recognition (ASR) models become ever more pervasive, it is important to ensure that they make reliable predictions under corruptions present in the physical and digital world. We propose Speech Robust Bench (SRB), a comprehensive benchmark for evaluating the robustness of ASR models to diverse corruptions. SRB is composed of 69 input perturbations which are intended to simulate various corruptions that ASR models may encounter in the physical and digital world. We use SRB to evaluate the robustness of several state-of-the-art ASR models and observe that model size and certain modeling choices such as discrete representations, and self-training appear to be conducive to robustness. We extend this analysis to measure the robustness of ASR models on data from various demographic subgroups, namely English and Spanish speakers, and males and females, and observed noticeable disparities in the model's robustness across subgroups. We believe that SRB will facilitate future research towards robust ASR models, by making it easier to conduct comprehensive and comparable robustness evaluations. △ Less

Submitted 8 March, 2024; originally announced March 2024.