@inproceedings{lazaridis-etal-2016-investigating,
title = "Investigating Cross-lingual Multi-level Adaptive Networks: The Importance of the Correlation of Source and Target Languages",
author = "Lazaridis, Alexandros and
Himawan, Ivan and
Motlicek, Petr and
Mporas, Iosif and
Garner, Philip N.",
editor = {Cettolo, Mauro and
Niehues, Jan and
St{\"u}ker, Sebastian and
Bentivogli, Luisa and
Cattoni, Rolando and
Federico, Marcello},
booktitle = "Proceedings of the 13th International Conference on Spoken Language Translation",
month = dec # " 8-9",
year = "2016",
address = "Seattle, Washington D.C",
publisher = "International Workshop on Spoken Language Translation",
url = "https://aclanthology.org/2016.iwslt-1.8/",
abstract = "The multi-level adaptive networks (MLAN) technique is a cross-lingual adaptation framework where a bottleneck (BN) layer in a deep neural network (DNN) trained in a source language is used for producing BN features to be exploited in a second DNN in a target language. We investigate how the correlation (in the sense of phonetic similarity) of the source and target languages and the amount of data of the source language affect the efficiency of the MLAN schemes. We experiment with three different scenarios using, i) French, as a source language uncorrelated to the target language, ii) Ukrainian, as a source language correlated to the target one and finally iii) English as a source language uncorrelated to the target language using a relatively large amount of data in respect to the other two scenarios. In all cases Russian is used as target language. GLOBALPHONE data is used, except for English, where a mixture of LIBRISPEECH, TEDLIUM and AMIDA is available. The results have shown that both of these two factors are important for the MLAN schemes. Specifically, on the one hand, when a modest amount of data from the source language is used, the correlation of the source and target languages is very important. On the other hand, the correlation of the two languages seems to be less important when a relatively large amount of data, from the source language, is used. The best performance in word error rate (WER), was achieved when the English language was used as the source one in the multi-task MLAN scheme, achieving a relative improvement of 9.4{\%} in respect to the baseline DNN model."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lazaridis-etal-2016-investigating">
<titleInfo>
<title>Investigating Cross-lingual Multi-level Adaptive Networks: The Importance of the Correlation of Source and Target Languages</title>
</titleInfo>
<name type="personal">
<namePart type="given">Alexandros</namePart>
<namePart type="family">Lazaridis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ivan</namePart>
<namePart type="family">Himawan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Petr</namePart>
<namePart type="family">Motlicek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Iosif</namePart>
<namePart type="family">Mporas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Philip</namePart>
<namePart type="given">N</namePart>
<namePart type="family">Garner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2016-dec 8-9</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 13th International Conference on Spoken Language Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mauro</namePart>
<namePart type="family">Cettolo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jan</namePart>
<namePart type="family">Niehues</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Stüker</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Luisa</namePart>
<namePart type="family">Bentivogli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rolando</namePart>
<namePart type="family">Cattoni</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marcello</namePart>
<namePart type="family">Federico</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>International Workshop on Spoken Language Translation</publisher>
<place>
<placeTerm type="text">Seattle, Washington D.C</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>The multi-level adaptive networks (MLAN) technique is a cross-lingual adaptation framework where a bottleneck (BN) layer in a deep neural network (DNN) trained in a source language is used for producing BN features to be exploited in a second DNN in a target language. We investigate how the correlation (in the sense of phonetic similarity) of the source and target languages and the amount of data of the source language affect the efficiency of the MLAN schemes. We experiment with three different scenarios using, i) French, as a source language uncorrelated to the target language, ii) Ukrainian, as a source language correlated to the target one and finally iii) English as a source language uncorrelated to the target language using a relatively large amount of data in respect to the other two scenarios. In all cases Russian is used as target language. GLOBALPHONE data is used, except for English, where a mixture of LIBRISPEECH, TEDLIUM and AMIDA is available. The results have shown that both of these two factors are important for the MLAN schemes. Specifically, on the one hand, when a modest amount of data from the source language is used, the correlation of the source and target languages is very important. On the other hand, the correlation of the two languages seems to be less important when a relatively large amount of data, from the source language, is used. The best performance in word error rate (WER), was achieved when the English language was used as the source one in the multi-task MLAN scheme, achieving a relative improvement of 9.4% in respect to the baseline DNN model.</abstract>
<identifier type="citekey">lazaridis-etal-2016-investigating</identifier>
<location>
<url>https://aclanthology.org/2016.iwslt-1.8/</url>
</location>
<part>
<date>2016-dec 8-9</date>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Investigating Cross-lingual Multi-level Adaptive Networks: The Importance of the Correlation of Source and Target Languages
%A Lazaridis, Alexandros
%A Himawan, Ivan
%A Motlicek, Petr
%A Mporas, Iosif
%A Garner, Philip N.
%Y Cettolo, Mauro
%Y Niehues, Jan
%Y Stüker, Sebastian
%Y Bentivogli, Luisa
%Y Cattoni, Rolando
%Y Federico, Marcello
%S Proceedings of the 13th International Conference on Spoken Language Translation
%D 2016
%8 dec 8 9
%I International Workshop on Spoken Language Translation
%C Seattle, Washington D.C
%F lazaridis-etal-2016-investigating
%X The multi-level adaptive networks (MLAN) technique is a cross-lingual adaptation framework where a bottleneck (BN) layer in a deep neural network (DNN) trained in a source language is used for producing BN features to be exploited in a second DNN in a target language. We investigate how the correlation (in the sense of phonetic similarity) of the source and target languages and the amount of data of the source language affect the efficiency of the MLAN schemes. We experiment with three different scenarios using, i) French, as a source language uncorrelated to the target language, ii) Ukrainian, as a source language correlated to the target one and finally iii) English as a source language uncorrelated to the target language using a relatively large amount of data in respect to the other two scenarios. In all cases Russian is used as target language. GLOBALPHONE data is used, except for English, where a mixture of LIBRISPEECH, TEDLIUM and AMIDA is available. The results have shown that both of these two factors are important for the MLAN schemes. Specifically, on the one hand, when a modest amount of data from the source language is used, the correlation of the source and target languages is very important. On the other hand, the correlation of the two languages seems to be less important when a relatively large amount of data, from the source language, is used. The best performance in word error rate (WER), was achieved when the English language was used as the source one in the multi-task MLAN scheme, achieving a relative improvement of 9.4% in respect to the baseline DNN model.
%U https://aclanthology.org/2016.iwslt-1.8/
Markdown (Informal)
[Investigating Cross-lingual Multi-level Adaptive Networks: The Importance of the Correlation of Source and Target Languages](https://aclanthology.org/2016.iwslt-1.8/) (Lazaridis et al., IWSLT 2016)
ACL