Data-Efficient ASR Personalization for Non-Normative Speech Using an Uncertainty-Based Phoneme Difficulty Score for Guided Sampling

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Pokel, Niclas, Moure, Pehuén, Böhringer, Roman, Gao, Yingqiang
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915866290946048
author Pokel, Niclas
Moure, Pehuén
Böhringer, Roman
Gao, Yingqiang
author_facet Pokel, Niclas
Moure, Pehuén
Böhringer, Roman
Gao, Yingqiang
contents ASR systems struggle with non-normative speech due to high acoustic variability and data scarcity. We propose a data-efficient method using phoneme-level uncertainty to guide fine-tuning for personalization. Instead of computationally expensive ensembles, we leverage Variational Low-Rank Adaptation (VI LoRA) to estimate epistemic uncertainty in foundation models. These estimates form a composite Phoneme Difficulty Score (PhDScore) that drives a targeted oversampling strategy. Evaluated on English and German datasets, including a longitudinal analysis against two clinical reports taken one year apart, we demonstrate that: (1) VI LoRA-based uncertainty aligns better with expert clinical assessments than standard entropy; (2) PhDScore captures stable, persistent articulatory difficulties; and (3) uncertainty-guided sampling significantly improves ASR accuracy for impaired speech.
format Preprint
id arxiv_https___arxiv_org_abs_2509_20396
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Data-Efficient ASR Personalization for Non-Normative Speech Using an Uncertainty-Based Phoneme Difficulty Score for Guided Sampling
Pokel, Niclas
Moure, Pehuén
Böhringer, Roman
Gao, Yingqiang
Audio and Speech Processing
Artificial Intelligence
Sound
ASR systems struggle with non-normative speech due to high acoustic variability and data scarcity. We propose a data-efficient method using phoneme-level uncertainty to guide fine-tuning for personalization. Instead of computationally expensive ensembles, we leverage Variational Low-Rank Adaptation (VI LoRA) to estimate epistemic uncertainty in foundation models. These estimates form a composite Phoneme Difficulty Score (PhDScore) that drives a targeted oversampling strategy. Evaluated on English and German datasets, including a longitudinal analysis against two clinical reports taken one year apart, we demonstrate that: (1) VI LoRA-based uncertainty aligns better with expert clinical assessments than standard entropy; (2) PhDScore captures stable, persistent articulatory difficulties; and (3) uncertainty-guided sampling significantly improves ASR accuracy for impaired speech.
title Data-Efficient ASR Personalization for Non-Normative Speech Using an Uncertainty-Based Phoneme Difficulty Score for Guided Sampling
topic Audio and Speech Processing
Artificial Intelligence
Sound
url https://arxiv.org/abs/2509.20396