The NAIST English speech recognition system for IWSLT 2013

Sakriani Sakti, Keigo Kubo, Graham Neubig, Tomoki Toda, Satoshi Nakamura


Abstract
This paper describes the NAIST English speech recognition system for the IWSLT 2013 Evaluation Campaign. In particular, we participated in the ASR track of the IWSLT TED task. Last year, we participated in collaboration with Karlsruhe Institute of Technology (KIT). This year is our first time to build a full-fledged ASR system for IWSLT solely developed by NAIST. Our final system utilizes weighted finitestate transducers with four-gram language models. The hypothesis selection is based on the principle of system combination. On the IWSLT official test set our system introduced in this work achieves a WER of 9.1% for tst2011, 10.0% for tst2012, and 16.2% for the new tst2013.
Anthology ID:
2013.iwslt-evaluation.23
Volume:
Proceedings of the 10th International Workshop on Spoken Language Translation: Evaluation Campaign
Month:
December 5-6
Year:
2013
Address:
Heidelberg, Germany
Editor:
Joy Ying Zhang
Venue:
IWSLT
SIG:
SIGSLT
Publisher:
Note:
Pages:
Language:
URL:
https://aclanthology.org/2013.iwslt-evaluation.23
DOI:
Bibkey:
Cite (ACL):
Sakriani Sakti, Keigo Kubo, Graham Neubig, Tomoki Toda, and Satoshi Nakamura. 2013. The NAIST English speech recognition system for IWSLT 2013. In Proceedings of the 10th International Workshop on Spoken Language Translation: Evaluation Campaign, Heidelberg, Germany.
Cite (Informal):
The NAIST English speech recognition system for IWSLT 2013 (Sakti et al., IWSLT 2013)
Copy Citation:
PDF:
https://aclanthology.org/2013.iwslt-evaluation.23.pdf