The MALACH corpus is packaged with benchmark train/test partitions and a 21.7% WER baseline using LSTM acoustic and language models.
There were many ways in which the 784 interviews could have been divided into training and test data
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Challenging the Boundaries of Speech Recognition: The MALACH Corpus
The MALACH corpus is packaged with benchmark train/test partitions and a 21.7% WER baseline using LSTM acoustic and language models.