Two-step phoneme-based ASR with LLM-P2G decoding, using noisy-phoneme augmentation and randomized top-K marginalized training, reduces WER on Polish and German versus WFST decoding.
These prior works share a similar motivation with ours that phoneme-based supervision is advantageous for multilingual acoustic representation learning
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.SD 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
LLM-based phoneme-to-grapheme for phoneme-based speech recognition
Two-step phoneme-based ASR with LLM-P2G decoding, using noisy-phoneme augmentation and randomized top-K marginalized training, reduces WER on Polish and German versus WFST decoding.