A streaming inverse text normalization model for Vietnamese combining PhoBERT tagging with dynamic chunk-size masking and a right-context buffer achieves F1 0.78, close to the non-streaming baseline's 0.86.
By modifying the architecture of the pretrained model, we enable real-time processing while taking advantage of the pretrained weights
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
unclear 1representative citing papers
citing papers explorer
-
Dynamic Context-Aware Streaming Pretrained Language Model For Inverse Text Normalization
A streaming inverse text normalization model for Vietnamese combining PhoBERT tagging with dynamic chunk-size masking and a right-context buffer achieves F1 0.78, close to the non-streaming baseline's 0.86.