This paper describes our English Speech-to-Text (STT) systems for the 2012 IWSLT TED ASR track evaluation. The systems consist of 10 subsystems that are combinations of different front-ends, e.g. MVDR based and MFCC based ones, and two different phone sets. The outputs of the subsystems are combined via confusion network combination. Decoding is done in two stages, where the systems of the second stage are adapted in an unsupervised manner on the combination of the first stage outputs using VTLN, MLLR, and cMLLR.
Index Terms: speech recognition, IWSLT, TED talks, evaluation system, system development
Cite as: Saam, C., Mohr, C., Kilgour, K., Heck, M., Sperber, M., Kubo, K., Stüker, S., Sakti, S., Neubig, G., Toda, T., Nakamura, S., Waibel, A. (2012) The 2012 KIT and KIT-NAIST English ASR systems for the IWSLT evaluation. Proc. International Workshop on Spoken Language Translation (IWSLT 2012), 87-90
@inproceedings{saam12_iwslt,
author={Christian Saam and Christian Mohr and Kevin Kilgour and Michael Heck and Matthias Sperber and Keigo Kubo and Sebastian Stüker and Sakriani Sakti and Graham Neubig and Tomoki Toda and Satoshi Nakamura and Alex Waibel},
title={{The 2012 KIT and KIT-NAIST English ASR systems for the IWSLT evaluation}},
year=2012,
booktitle={Proc. International Workshop on Spoken Language Translation (IWSLT 2012)},
pages={87--90}
}