Towards a Single ASR Model That Generalizes to Disordered Speech

TL;DR

Integrating a small amount of disordered speech data into a state-of-the-art ASR model significantly improves recognition accuracy for disordered speech without harming standard benchmarks, advancing accessible speech technology.

Contribution

This work demonstrates that a small dataset of disordered speech can substantially enhance ASR performance on disordered speech, bridging the gap with personalized models.

Findings

33% improvement on prompted disordered speech

26% improvement on spontaneous disordered speech

No significant decline on standard benchmarks

Abstract

This study investigates the impact of integrating a dataset of disordered speech recordings ($\sim$1,000 hours) into the fine-tuning of a near state-of-the-art ASR baseline system. Contrary to what one might expect, despite the data being less than 1% of the training data of the ASR system, we find a considerable improvement in disordered speech recognition accuracy. Specifically, we observe a 33% improvement on prompted speech, and a 26% improvement on a newly gathered spontaneous, conversational dataset of disordered speech. Importantly, there is no significant performance decline on standard speech recognition benchmarks. Further, we observe that the proposed tuning strategy helps close the gap between the baseline system and personalized models by 64% highlighting the significant progress as well as the room for improvement. Given the substantial benefits of our findings, this experiment suggests that from a fairness perspective, incorporating a small fraction of high quality disordered speech data in a training recipe is an easy step that could be done to make speech technology more accessible for users with speech disabilities.