Hypernetworks for Personalizing ASR to Atypical Speech
This work addresses the challenge of adapting ASR models for individuals with diverse atypical speech characteristics, where data scarcity and variability limit traditional methods, offering a more accessible and efficient solution.
The paper tackled the problem of personalizing automatic speech recognition (ASR) for atypical speech without requiring prior knowledge of specific disorders, by using a meta-learned hypernetwork to generate individualized adaptations on-the-fly, achieving a 75.2% relative reduction in Word Error Rate (WER) with only 0.1% of the full parameter budget.
Parameter-efficient fine-tuning (PEFT) for personalizing automatic speech recognition (ASR) has recently shown promise for adapting general population models to atypical speech. However, these approaches assume a priori knowledge of the atypical speech disorder being adapted for -- the diagnosis of which requires expert knowledge that is not always available. Even given this knowledge, data scarcity and high inter/intra-speaker variability further limit the effectiveness of traditional fine-tuning. To circumvent these challenges, we first identify the minimal set of model parameters required for ASR adaptation. Our analysis of each individual parameter's effect on adaptation performance allows us to reduce Word Error Rate (WER) by half while adapting 0.03% of all weights. Alleviating the need for cohort-specific models, we next propose the novel use of a meta-learned hypernetwork to generate highly individualized, utterance-level adaptations on-the-fly for a diverse set of atypical speech characteristics. Evaluating adaptation at the global, cohort and individual-level, we show that hypernetworks generalize better to out-of-distribution speakers, while maintaining an overall relative WER reduction of 75.2% using 0.1% of the full parameter budget.