This paper describes our automatic speech recognition (ASR) systemfor the IWSLT 2012 evaluation campaign. The target data of the campaign is selected from the TED talks, a collection of public speeches on a variety of topics spoken in English. Our ASR system is based on weighted finite-state transducers and exploits an combination of acoustic models for spontaneous speech, language models based on ngram and factored recurrent neural network trained with effectively selected corpora, and unsupervised topic adaptation framework utilizing ASR results. Accordingly, the system achieved 10.6% and 12.0% word error rate for the tst2011 and tst2012 evaluation set, respectively.
Cite as: Yamamoto, H., Wu, Y., Huang, C.-L., Lu, X., Dixon, P.R., Matsuda, S., Hori, C., Kashioka, H. (2012) The NICT ASR system for IWSLT2012. Proc. International Workshop on Spoken Language Translation (IWSLT 2012), 34-37
@inproceedings{yamamoto12_iwslt,
author={Hitoshi Yamamoto and Youzheng Wu and Chien-Lin Huang and Xugang Lu and Paul R. Dixon and Shigeki Matsuda and Chiori Hori and Hideki Kashioka},
title={{The NICT ASR system for IWSLT2012}},
year=2012,
booktitle={Proc. International Workshop on Spoken Language Translation (IWSLT 2012)},
pages={34--37}
}