@inproceedings{ding-etal-2022-redistributing,
title = "Redistributing Low-Frequency Words: Making the Most of Monolingual Data in Non-Autoregressive Translation",
author = "Ding, Liang and
Wang, Longyue and
Shi, Shuming and
Tao, Dacheng and
Tu, Zhaopeng",
editor = "Muresan, Smaranda and
Nakov, Preslav and
Villavicencio, Aline",
booktitle = "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = may,
year = "2022",
address = "Dublin, Ireland",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.acl-long.172",
doi = "10.18653/v1/2022.acl-long.172",
pages = "2417--2426",
abstract = "Knowledge distillation (KD) is the preliminary step for training non-autoregressive translation (NAT) models, which eases the training of NAT models at the cost of losing important information for translating low-frequency words. In this work, we provide an appealing alternative for NAT {--} \textit{monolingual KD}, which trains NAT student on external monolingual data with AT teacher trained on the original bilingual data. Monolingual KD is able to transfer both the knowledge of the original bilingual data (implicitly encoded in the trained AT teacher model) and that of the new monolingual data to the NAT student model. Extensive experiments on eight WMT benchmarks over two advanced NAT models show that monolingual KD consistently outperforms the standard KD by improving low-frequency word translation, without introducing any computational cost. Monolingual KD enjoys desirable expandability, which can be further enhanced (when given more computational budget) by combining with the standard KD, a reverse monolingual KD, or enlarging the scale of monolingual data. Extensive analyses demonstrate that these techniques can be used together profitably to further recall the useful information lost in the standard KD. Encouragingly, combining with standard KD, our approach achieves 30.4 and 34.1 BLEU points on the WMT14 English-German and German-English datasets, respectively. Our code and trained models are freely available at \url{https://github.com/alphadl/RLFW-NAT.mono}.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ding-etal-2022-redistributing">
<titleInfo>
<title>Redistributing Low-Frequency Words: Making the Most of Monolingual Data in Non-Autoregressive Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Liang</namePart>
<namePart type="family">Ding</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Longyue</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuming</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dacheng</namePart>
<namePart type="family">Tao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhaopeng</namePart>
<namePart type="family">Tu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2022-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Smaranda</namePart>
<namePart type="family">Muresan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Preslav</namePart>
<namePart type="family">Nakov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aline</namePart>
<namePart type="family">Villavicencio</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Dublin, Ireland</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Knowledge distillation (KD) is the preliminary step for training non-autoregressive translation (NAT) models, which eases the training of NAT models at the cost of losing important information for translating low-frequency words. In this work, we provide an appealing alternative for NAT – monolingual KD, which trains NAT student on external monolingual data with AT teacher trained on the original bilingual data. Monolingual KD is able to transfer both the knowledge of the original bilingual data (implicitly encoded in the trained AT teacher model) and that of the new monolingual data to the NAT student model. Extensive experiments on eight WMT benchmarks over two advanced NAT models show that monolingual KD consistently outperforms the standard KD by improving low-frequency word translation, without introducing any computational cost. Monolingual KD enjoys desirable expandability, which can be further enhanced (when given more computational budget) by combining with the standard KD, a reverse monolingual KD, or enlarging the scale of monolingual data. Extensive analyses demonstrate that these techniques can be used together profitably to further recall the useful information lost in the standard KD. Encouragingly, combining with standard KD, our approach achieves 30.4 and 34.1 BLEU points on the WMT14 English-German and German-English datasets, respectively. Our code and trained models are freely available at https://github.com/alphadl/RLFW-NAT.mono.</abstract>
<identifier type="citekey">ding-etal-2022-redistributing</identifier>
<identifier type="doi">10.18653/v1/2022.acl-long.172</identifier>
<location>
<url>https://aclanthology.org/2022.acl-long.172</url>
</location>
<part>
<date>2022-05</date>
<extent unit="page">
<start>2417</start>
<end>2426</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Redistributing Low-Frequency Words: Making the Most of Monolingual Data in Non-Autoregressive Translation
%A Ding, Liang
%A Wang, Longyue
%A Shi, Shuming
%A Tao, Dacheng
%A Tu, Zhaopeng
%Y Muresan, Smaranda
%Y Nakov, Preslav
%Y Villavicencio, Aline
%S Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2022
%8 May
%I Association for Computational Linguistics
%C Dublin, Ireland
%F ding-etal-2022-redistributing
%X Knowledge distillation (KD) is the preliminary step for training non-autoregressive translation (NAT) models, which eases the training of NAT models at the cost of losing important information for translating low-frequency words. In this work, we provide an appealing alternative for NAT – monolingual KD, which trains NAT student on external monolingual data with AT teacher trained on the original bilingual data. Monolingual KD is able to transfer both the knowledge of the original bilingual data (implicitly encoded in the trained AT teacher model) and that of the new monolingual data to the NAT student model. Extensive experiments on eight WMT benchmarks over two advanced NAT models show that monolingual KD consistently outperforms the standard KD by improving low-frequency word translation, without introducing any computational cost. Monolingual KD enjoys desirable expandability, which can be further enhanced (when given more computational budget) by combining with the standard KD, a reverse monolingual KD, or enlarging the scale of monolingual data. Extensive analyses demonstrate that these techniques can be used together profitably to further recall the useful information lost in the standard KD. Encouragingly, combining with standard KD, our approach achieves 30.4 and 34.1 BLEU points on the WMT14 English-German and German-English datasets, respectively. Our code and trained models are freely available at https://github.com/alphadl/RLFW-NAT.mono.
%R 10.18653/v1/2022.acl-long.172
%U https://aclanthology.org/2022.acl-long.172
%U https://doi.org/10.18653/v1/2022.acl-long.172
%P 2417-2426
Markdown (Informal)
[Redistributing Low-Frequency Words: Making the Most of Monolingual Data in Non-Autoregressive Translation](https://aclanthology.org/2022.acl-long.172) (Ding et al., ACL 2022)
ACL