Jointly training pitch accent detection with ASR on wav2vec2 reduces LibriSpeech WER from 6.0 to 4.3 in a one-hour fine-tuning setting.
Datasets Our research uses the Boston University Radio News Corpus (BURNC) [7], Librispeech [45] and Libri-light [46] corpora
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
-
Pitch Accent Detection improves Pretrained Automatic Speech Recognition
Jointly training pitch accent detection with ASR on wav2vec2 reduces LibriSpeech WER from 6.0 to 4.3 in a one-hour fine-tuning setting.