On Dutch end-to-end ASR systems, average word error rate hides large performance gaps across speaker groups, and bias mitigation can lower the average while increasing bias.
Large datasets and advances in deep learning have significantly improved the per- formance of speech technologies [2, 3]
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
background 1representative citing papers
citing papers explorer
-
How to Evaluate Automatic Speech Recognition: Comparing Different Performance and Bias Measures
On Dutch end-to-end ASR systems, average word error rate hides large performance gaps across speaker groups, and bias mitigation can lower the average while increasing bias.