Improving the training and evaluation efficiency of recurrent neural network language models | IEEE Conference Publication | IEEE Xplore