Repository logo
 

Improving the training and evaluation efficiency of recurrent neural network language models


Type

Conference Object

Change log

Authors

Chen, X 
Liu, X 
Gales, MJF 
Woodland, PC 

Abstract

Recurrent neural network language models (RNNLMs) are becoming increasingly popular for speech recognition. Previously, we have shown that RNNLMs with a full (non-classed) output layer (F-RNNLMs) can be trained efficiently using a GPU giving a large reduction in training time over conventional class-based models (C-RNNLMs) on a standard CPU. However, since test-time RNNLM evaluation is often performed entirely on a CPU, standard F-RNNLMs are inefficient since the entire output layer needs to be calculated for normalisation. In this paper, it is demonstrated that C-RNNLMs can be efficiently trained on a GPU, using our spliced sentence bunch technique which allows good CPU test-time performance (42x speedup over F-RNNLM). Furthermore, the performance of different classing approaches is investigated. We also examine the use of variance regularisation of the softmax denominator for F-RNNLMs and show that it allows F-RNNLMs to be efficiently used in test (56x speedup on CPU). Finally the use of two GPUs for F-RNNLM training using pipelining is described and shown to give a reduction in training time over a single GPU by a factor of 1.6.

Description

Keywords

language models, recurrent neural network, GPU, speech recognition

Journal Title

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

Conference Name

ICASSP 2015 - 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Journal ISSN

1520-6149

Volume Title

Publisher

IEEE
Sponsorship
Xie Chen is supported by Toshiba Research Europe Ltd, Cambridge Research Lab. The research leading to these results was also supported by EPSRC grant EP/I031022/1 (Natural Speech Technology) and DARPA under the Broad Operational Language Translation (BOLT) and RATS programs. The paper does not necessarily reflect the position or the policy of US Government and no official endorsement should be inferred.