Acoustic-focused augmentation of a 960-hour dataset is reported to reduce out-of-distribution word error rates by up to 19.24 percent, suggesting acoustic diversity, not linguistic diversity, drives ASR robustness.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Towards Pretraining Robust ASR Foundation Model with Acoustic-Aware Data Augmentation
Acoustic-focused augmentation of a 960-hour dataset is reported to reduce out-of-distribution word error rates by up to 19.24 percent, suggesting acoustic diversity, not linguistic diversity, drives ASR robustness.