Figure 7From: RNN language model with word clustering and class-based output layer WER convergence of 100-best rescoring with RNNLM-Freq/Brown on Hub5’00-SWB set. Back to article page