Applying dropout during decoding to generate multiple candidate transcripts, then selecting confident ones, improves semi-supervised end-to-end ASR by 2% absolute WER on TEDLIUM.
Experiments are performed on TEDLIUM and Table 1: Training, adaptation and test data for different dataset
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Exploiting semi-supervised training through a dropout regularization in end-to-end speech recognition
Applying dropout during decoding to generate multiple candidate transcripts, then selecting confident ones, improves semi-supervised end-to-end ASR by 2% absolute WER on TEDLIUM.