Result Details

Combining Forward and Backward Search in Decoding

HANNEMANN, M.; POVEY, D.; ZWEIG, G. Combining Forward and Backward Search in Decoding. Proceedings of ICASSP 2013. Vancouver: IEEE Signal Processing Society, 2013. p. 6739-6743. ISBN: 978-1-4799-0355-9.
Type
conference paper
Language
English
Authors
Hannemann Mirko, Ph.D., DCGM (FIT)
Povey Daniel
Zweig Geoffrey
Abstract

This article describes a combination of forward and backward search in speech decoding based on WFST decoders.

Keywords

speech decoding, beam width, search errors

URL
Annotation

We introduce a speed-up for weighted finite state transducer (WFST) based decoders, which is based on the idea that one decoding pass using a wider beam can be replaced by two decoding passes with smaller beams, decoding forward and backward in time. We apply this in a decoder that works with a variable beam width, which is widened in areas where the two decoding passes disagree. Experimental results are shown on the Wall Street Journal corpus (WSJ) using the Kaldi toolkit, and show a substantial speedup (a factor or 2 or 3) at the "more accurate" operating points. As part of this work we also introduce a new fast algorithm for weight pushing in WFSTs, and summarize an algorithm for the time reversal of backoff language models.

Published
2013
Pages
6739–6743
Proceedings
Proceedings of ICASSP 2013
Conference
38th International Conference on Acoustics, Speech, and Signal Processing
ISBN
978-1-4799-0355-9
Publisher
IEEE Signal Processing Society
Place
Vancouver
BibTeX
@inproceedings{BUT103491,
  author="Mirko {Hannemann} and Daniel {Povey} and Geoffrey {Zweig}",
  title="Combining Forward and Backward Search in Decoding",
  booktitle="Proceedings of ICASSP 2013",
  year="2013",
  pages="6739--6743",
  publisher="IEEE Signal Processing Society",
  address="Vancouver",
  isbn="978-1-4799-0355-9",
  url="http://www.fit.vutbr.cz/research/groups/speech/publi/2013/hannemann_icassp2013_0006739.pdf"
}
Projects
Centrum excelence IT4Innovations, MŠMT, Operační program Výzkum a vývoj pro inovace, ED1.1.00/02.0070, start: 2011-01-01, end: 2015-12-31, completed
IARPA Building Speech Recognition for Keyword Search in a New Language in a Week with Limited Training Data (BABEL) - Babelon, BBN, start: 2012-03-05, end: 2016-11-04, completed
Security-Oriented Research in Information Technology, MŠMT, Institucionální prostředky SR ČR (např. VZ, VC), MSM0021630528, start: 2007-01-01, end: 2013-12-31, running
Research groups
Departments
Back to top