Cache based recurrent neural network language model inference for first pass speech recognition

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...

Full description

Saved in:
Bibliographic Details
Published in2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) pp. 6354 - 6358
Main Authors Zhiheng Huang, Zweig, Geoffrey, Dumoulin, Benoit
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.05.2014
Subjects
cache
computational efficiency
Computational modeling
Data models
Decoding
Hidden Markov models
History
recurrent neural network language model
Recurrent neural networks
Speech recognition
voice search
Online AccessGet full text

Cover

More Information
Summary: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.
ISSN:1520-6149
2379-190X
DOI:10.1109/ICASSP.2014.6854827