Weakly supervised training on utterance-level pronunciation labels followed by fine-tuning on a small set of selected phoneme-labeled utterances produces phoneme-level scoring performance comparable to models trained with full phoneme-level supervision.
These aspects can be evaluated at different granularities (phoneme, word, or sentence) and mod- eled separately or jointly [1, 2]
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
-
A study on weakly-supervised training approaches for phoneme-level pronunciation scoring
Weakly supervised training on utterance-level pronunciation labels followed by fine-tuning on a small set of selected phoneme-labeled utterances produces phoneme-level scoring performance comparable to models trained with full phoneme-level supervision.