Skip to main content
EURASIP Journal on Audio, Speech, and Music Processing

Table 4 WER for development set rescored with different RNNLMs and LM-KN5

From: RNN language model with word clustering and class-based output layer

 

Class

Model

50

100

200

400

LM-KN5 + RNNLM-Freq (%)

11.7

11.6

11.6

11.7

LM-KN5 + RNNLM-Brown (%)

11.5

115̇

11.5

11.4

  1. The WER of 1-best hypothesis is 13.4% and the WER for LM-KN5 is 12.9%.
Back to article page