wav2vec2.0 models classify stop burst presence with 88 to 94 percent accuracy, and a Japanese-pretrained model reaches near-peak accuracy on English with only 500 annotated tokens.
Specifically, we are interested in whether such models can predict the presence or absence of a stop burst in datasets reflecting two broad types of collected speech data: 1
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
unclear 1representative citing papers
citing papers explorer
-
Automatic classification of stop realisation with wav2vec2.0
wav2vec2.0 models classify stop burst presence with 88 to 94 percent accuracy, and a Japanese-pretrained model reaches near-peak accuracy on English with only 500 annotated tokens.