Introduces truncation and Jacobian-based measures of effective context showing that self-supervised speech Transformers use a short, mostly local context and can be streamed with minor probing degradation.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Effective Context in Neural Speech Models
Introduces truncation and Jacobian-based measures of effective context showing that self-supervised speech Transformers use a short, mostly local context and can be streamed with minor probing degradation.