A systematic child ASR benchmark shows adult-trained SSL features are biased against child speech, flat-start training on child data helps, and zero-shot scaling plateaus near 1B parameters.
Additional support was provided by FWO-SBO grant S004923N: NEFL, KU Leuven C24M/22/025, and FWO grant V401325N
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Benchmarking Training Paradigms, Dataset Composition, and Model Scaling for Child ASR in ESPnet
A systematic child ASR benchmark shows adult-trained SSL features are biased against child speech, flat-start training on child data helps, and zero-shot scaling plateaus near 1B parameters.