Utterance-level selection methods identify reliable ASR outputs for child speech with precision above 97.4 percent, enabling 21 to 55.9 percent of read and dialogue datasets to be retained with utterance error rates below 2.6 percent across English and Dutch.
Adaptation of whisper models to child speech recognition,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Utterance-Level Methods for Identifying Reliable ASR-Output for Child Speech
Utterance-level selection methods identify reliable ASR outputs for child speech with precision above 97.4 percent, enabling 21 to 55.9 percent of read and dialogue datasets to be retained with utterance error rates below 2.6 percent across English and Dutch.