An 8B LLaMa decoder with a Conformer audio encoder generates disfluency tokens and timestamps, and works even when the text hints come from imperfect phoneme or word aligners.
Smooth- LLaMa leverages contextual understanding to refine noisy data, achieving precise outputs even with imperfect textual inputs, such as word- or phoneme-level alignments
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
-
Smooth Operators: LLMs Translating Imperfect Hints into Disfluency-Rich Transcripts
An 8B LLaMa decoder with a Conformer audio encoder generates disfluency tokens and timestamps, and works even when the text hints come from imperfect phoneme or word aligners.