Multitask G2P model with built-in Anglicism classifier improves pronunciation handling and reduces error rates by 1-3% in German ASR on a dedicated test set.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2021 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Multitask Learning for Grapheme-to-Phoneme Conversion of Anglicisms in German Speech Recognition
Multitask G2P model with built-in Anglicism classifier improves pronunciation handling and reduces error rates by 1-3% in German ASR on a dedicated test set.