For a fixed training budget, low-resource ASR models handle unseen accents best when trained on more speakers with less audio each, and accent diversity in training gives minimal extra benefit.
In Figure 1, 2We observe strong correlation between CER and Word Error Rate (WER) (r= 0.92) in our evaluation, so we report CER only for brevity
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Effects of Speaker Count, Duration, and Accent Diversity on Zero-Shot Accent Robustness in Low-Resource ASR
For a fixed training budget, low-resource ASR models handle unseen accents best when trained on more speakers with less audio each, and accent diversity in training gives minimal extra benefit.