Using word lattice information for a tighter coupling in speech translation systems

Shirin Saleem, Szu Chen Jou, Stephan Vogel, Tanja Schultz

Research output: Chapter in Book/Report/Conference proceedingConference contribution

22 Citations (Scopus)

Abstract

In this paper we present first experiments towards a tighter coupling between Automatic Speech Recognition (ASR) and Statistical Machine Translation (SMT) to improve the overall performance of our speech translation system. In coventional speech translation systems, the recognizer outputs a single hypothesis which is then translated by the SMT system. This approach has the limitation of being largely dependent on the word error rate of the first best hypothesis. The word error rate is typically lowered by generating many alternative hypotheses in the form of a word lattice. The information in the word lattice and the scores from the recognizer can be used by the translation system to obtain better performance. In our experiments, by switching from the single best hypotheses to word lattices as the interface between ASR and SMT, and by introducing weighted acoustic scores in the translation system, the overall performance was increased by 16.22%.

Original languageEnglish
Title of host publication8th International Conference on Spoken Language Processing, ICSLP 2004
PublisherInternational Speech Communication Association
Pages41-44
Number of pages4
Publication statusPublished - 2004
Externally publishedYes
Event8th International Conference on Spoken Language Processing, ICSLP 2004 - Jeju, Jeju Island, Korea, Republic of
Duration: 4 Oct 20048 Oct 2004

Other

Other8th International Conference on Spoken Language Processing, ICSLP 2004
CountryKorea, Republic of
CityJeju, Jeju Island
Period4/10/048/10/04

Fingerprint

performance
experiment
Translation System
acoustics
Statistical Machine Translation
Automatic Speech Recognition
Experiment
Acoustics
Machine Translation System

ASJC Scopus subject areas

  • Language and Linguistics
  • Linguistics and Language

Cite this

Saleem, S., Jou, S. C., Vogel, S., & Schultz, T. (2004). Using word lattice information for a tighter coupling in speech translation systems. In 8th International Conference on Spoken Language Processing, ICSLP 2004 (pp. 41-44). International Speech Communication Association.

Using word lattice information for a tighter coupling in speech translation systems. / Saleem, Shirin; Jou, Szu Chen; Vogel, Stephan; Schultz, Tanja.

8th International Conference on Spoken Language Processing, ICSLP 2004. International Speech Communication Association, 2004. p. 41-44.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Saleem, S, Jou, SC, Vogel, S & Schultz, T 2004, Using word lattice information for a tighter coupling in speech translation systems. in 8th International Conference on Spoken Language Processing, ICSLP 2004. International Speech Communication Association, pp. 41-44, 8th International Conference on Spoken Language Processing, ICSLP 2004, Jeju, Jeju Island, Korea, Republic of, 4/10/04.
Saleem S, Jou SC, Vogel S, Schultz T. Using word lattice information for a tighter coupling in speech translation systems. In 8th International Conference on Spoken Language Processing, ICSLP 2004. International Speech Communication Association. 2004. p. 41-44
Saleem, Shirin ; Jou, Szu Chen ; Vogel, Stephan ; Schultz, Tanja. / Using word lattice information for a tighter coupling in speech translation systems. 8th International Conference on Spoken Language Processing, ICSLP 2004. International Speech Communication Association, 2004. pp. 41-44
@inproceedings{c09d3398df8c4c6e8a8fdf0ee64a561b,
title = "Using word lattice information for a tighter coupling in speech translation systems",
abstract = "In this paper we present first experiments towards a tighter coupling between Automatic Speech Recognition (ASR) and Statistical Machine Translation (SMT) to improve the overall performance of our speech translation system. In coventional speech translation systems, the recognizer outputs a single hypothesis which is then translated by the SMT system. This approach has the limitation of being largely dependent on the word error rate of the first best hypothesis. The word error rate is typically lowered by generating many alternative hypotheses in the form of a word lattice. The information in the word lattice and the scores from the recognizer can be used by the translation system to obtain better performance. In our experiments, by switching from the single best hypotheses to word lattices as the interface between ASR and SMT, and by introducing weighted acoustic scores in the translation system, the overall performance was increased by 16.22{\%}.",
author = "Shirin Saleem and Jou, {Szu Chen} and Stephan Vogel and Tanja Schultz",
year = "2004",
language = "English",
pages = "41--44",
booktitle = "8th International Conference on Spoken Language Processing, ICSLP 2004",
publisher = "International Speech Communication Association",

}

TY - GEN

T1 - Using word lattice information for a tighter coupling in speech translation systems

AU - Saleem, Shirin

AU - Jou, Szu Chen

AU - Vogel, Stephan

AU - Schultz, Tanja

PY - 2004

Y1 - 2004

N2 - In this paper we present first experiments towards a tighter coupling between Automatic Speech Recognition (ASR) and Statistical Machine Translation (SMT) to improve the overall performance of our speech translation system. In coventional speech translation systems, the recognizer outputs a single hypothesis which is then translated by the SMT system. This approach has the limitation of being largely dependent on the word error rate of the first best hypothesis. The word error rate is typically lowered by generating many alternative hypotheses in the form of a word lattice. The information in the word lattice and the scores from the recognizer can be used by the translation system to obtain better performance. In our experiments, by switching from the single best hypotheses to word lattices as the interface between ASR and SMT, and by introducing weighted acoustic scores in the translation system, the overall performance was increased by 16.22%.

AB - In this paper we present first experiments towards a tighter coupling between Automatic Speech Recognition (ASR) and Statistical Machine Translation (SMT) to improve the overall performance of our speech translation system. In coventional speech translation systems, the recognizer outputs a single hypothesis which is then translated by the SMT system. This approach has the limitation of being largely dependent on the word error rate of the first best hypothesis. The word error rate is typically lowered by generating many alternative hypotheses in the form of a word lattice. The information in the word lattice and the scores from the recognizer can be used by the translation system to obtain better performance. In our experiments, by switching from the single best hypotheses to word lattices as the interface between ASR and SMT, and by introducing weighted acoustic scores in the translation system, the overall performance was increased by 16.22%.

UR - http://www.scopus.com/inward/record.url?scp=85009064092&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85009064092&partnerID=8YFLogxK

M3 - Conference contribution

SP - 41

EP - 44

BT - 8th International Conference on Spoken Language Processing, ICSLP 2004

PB - International Speech Communication Association

ER -