Publication Details
Promising Accurate Prefix Boosting For Sequence-to-sequence ASR
BASKAR, M.; BURGET, L.; WATANABE, S.; KARAFIÁT, M.; HORI, T.; ČERNOCKÝ, J. Promising Accurate Prefix Boosting For Sequence-to-sequence ASR. In Proceedings of ICASSP. Brighton: IEEE Signal Processing Society, 2019. p. 5646-5650. ISBN: 978-1-5386-4658-8.
Czech title
Slibná technika pro přesné zdůrazňování prexixů pro ASR založené na převodu sekvencí na sekvence
Type
conference paper
Language
English
Authors
Baskar Murali Karthick, Ing., Ph.D.
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Watanabe Shinji (FIT)
Karafiát Martin, Ing., Ph.D. (DCGM)
HORI, T.
Černocký Jan, prof. Dr. Ing. (DCGM)
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Watanabe Shinji (FIT)
Karafiát Martin, Ing., Ph.D. (DCGM)
HORI, T.
Černocký Jan, prof. Dr. Ing. (DCGM)
URL
Keywords
Beam search training, sequence learning, discriminative training, Attention models, softmax-margin
Abstract
In this paper, we present promising accurate prefix boosting (PAPB), a discriminative training technique for attention based sequence-tosequence (seq2seq) ASR. PAPB is devised to unify the training and testing scheme effectively. The training procedure involves maximizing the score of each partial correct sequence obtained during beam search compared to other hypotheses. The training objective also includes minimization of token (character) error rate. PAPB shows its efficacy by achieving 10.8% and 3.8% WER with and without external RNNLM respectively on Wall Street Journal dataset.
Published
2019
Pages
5646–5650
Proceedings
Proceedings of ICASSP
ISBN
978-1-5386-4658-8
Publisher
IEEE Signal Processing Society
Place
Brighton
DOI
UT WoS
000482554005176
EID Scopus
BibTeX
@inproceedings{BUT160001,
author="BASKAR, M. and BURGET, L. and WATANABE, S. and KARAFIÁT, M. and HORI, T. and ČERNOCKÝ, J.",
title="Promising Accurate Prefix Boosting For Sequence-to-sequence ASR",
booktitle="Proceedings of ICASSP",
year="2019",
pages="5646--5650",
publisher="IEEE Signal Processing Society",
address="Brighton",
doi="10.1109/ICASSP.2019.8682782",
isbn="978-1-5386-4658-8",
url="https://ieeexplore.ieee.org/document/8682782"
}