Zhiheng Huang, Geoffrey Zweig, and Benoit Dumoulin
Recurrent neural network language models (RNNLMs) have recently produced improvements on language processing tasks ranging from machine translation to word tagging and speech recognition. To date, however, the computational expense of RNNLMs has hampered their application to first pass decoding. In this paper, we show that by restricting the RNNLM calls to those words that receive a reasonable score according to a n-gram model, and by deploying a set of caches, we can reduce the cost of using an RNNLM in the first pass to that of using an additional n-gram model. We compare this scheme to lattice rescoring, and find that they produce comparable results for a Bing Voice search task. The best performance results from rescoring a lattice that is itself created with a RNNLM in the first pass.
|Publisher||IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)|