@inproceedings{somers-sugita-2003-evaluating,
title = "Evaluating commercial spoken language translation software",
author = "Somers, Harold and
Sugita, Yuri",
booktitle = "Proceedings of Machine Translation Summit IX: Papers",
month = sep # " 23-27",
year = "2003",
address = "New Orleans, USA",
url = "https://aclanthology.org/2003.mtsummit-papers.49",
abstract = "While spoken language translation remains a research goal, a crude form of it is widely available commercially for Japanese{--}English as a pipeline concatenation of speech-to-text recognition (SR), text-to-text translation (MT) and text-to-speech synthesis (SS). This paper proposes and illustrates an evaluation methodology for this noisy channel which tries to quantify the relative amount of degradation in translation quality due to each of the contributing modules. A small pilot experiment involving word-accuracy rate for the SR, and a fidelity evaluation for the MT and SS modules is proposed in which subjects are asked to paraphrase translated and/or synthesised sentences from a tourist{'}s phrasebook. Results show (as expected) that MT is the {``}noisiest{''} channel, with SS contributing least noise. The concatenation of the three channels is worse than could be predicted from the performance of each as individual tasks.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="somers-sugita-2003-evaluating">
<titleInfo>
<title>Evaluating commercial spoken language translation software</title>
</titleInfo>
<name type="personal">
<namePart type="given">Harold</namePart>
<namePart type="family">Somers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuri</namePart>
<namePart type="family">Sugita</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2003-sep" 23-27"</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of Machine Translation Summit IX: Papers</title>
</titleInfo>
<originInfo>
<place>
<placeTerm type="text">New Orleans, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>While spoken language translation remains a research goal, a crude form of it is widely available commercially for Japanese–English as a pipeline concatenation of speech-to-text recognition (SR), text-to-text translation (MT) and text-to-speech synthesis (SS). This paper proposes and illustrates an evaluation methodology for this noisy channel which tries to quantify the relative amount of degradation in translation quality due to each of the contributing modules. A small pilot experiment involving word-accuracy rate for the SR, and a fidelity evaluation for the MT and SS modules is proposed in which subjects are asked to paraphrase translated and/or synthesised sentences from a tourist’s phrasebook. Results show (as expected) that MT is the “noisiest” channel, with SS contributing least noise. The concatenation of the three channels is worse than could be predicted from the performance of each as individual tasks.</abstract>
<identifier type="citekey">somers-sugita-2003-evaluating</identifier>
<location>
<url>https://aclanthology.org/2003.mtsummit-papers.49</url>
</location>
<part>
<date>2003-sep" 23-27"</date>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Evaluating commercial spoken language translation software
%A Somers, Harold
%A Sugita, Yuri
%S Proceedings of Machine Translation Summit IX: Papers
%D 2003
%8 sep" 23 27"
%C New Orleans, USA
%F somers-sugita-2003-evaluating
%X While spoken language translation remains a research goal, a crude form of it is widely available commercially for Japanese–English as a pipeline concatenation of speech-to-text recognition (SR), text-to-text translation (MT) and text-to-speech synthesis (SS). This paper proposes and illustrates an evaluation methodology for this noisy channel which tries to quantify the relative amount of degradation in translation quality due to each of the contributing modules. A small pilot experiment involving word-accuracy rate for the SR, and a fidelity evaluation for the MT and SS modules is proposed in which subjects are asked to paraphrase translated and/or synthesised sentences from a tourist’s phrasebook. Results show (as expected) that MT is the “noisiest” channel, with SS contributing least noise. The concatenation of the three channels is worse than could be predicted from the performance of each as individual tasks.
%U https://aclanthology.org/2003.mtsummit-papers.49
Markdown (Informal)
[Evaluating commercial spoken language translation software](https://aclanthology.org/2003.mtsummit-papers.49) (Somers & Sugita, MTSummit 2003)
ACL