@inproceedings{yuan-etal-2023-lego,
title = "{L}ego-{MT}: Learning Detachable Models for Massively Multilingual Machine Translation",
author = "Yuan, Fei and
Lu, Yinquan and
Zhu, Wenhao and
Kong, Lingpeng and
Li, Lei and
Qiao, Yu and
Xu, Jingjing",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-acl.731",
doi = "10.18653/v1/2023.findings-acl.731",
pages = "11518--11533",
abstract = "Multilingual neural machine translation (MNMT) aims to build a unified model for many language directions. Existing monolithic models for MNMT encounter two challenges: parameter interference among languages and inefficient inference for large models. In this paper, we revisit the classic multi-way structures and develop a detachable model by assigning each language (or group of languages) to an individual branch that supports plug-and-play training and inference. To address the needs of learning representations for all languages in a unified space, we propose a novel efficient training recipe, upon which we build an effective detachable model, Lego-MT.For a fair comparison, we collect data from OPUS and build a translation benchmark covering 433 languages and 1.3B parallel data. Experiments show that Lego-MT with 1.2B parameters brings an average gain of 3.2 spBLEU. It even outperforms M2M-100 with 12B parameters. The proposed training recipe brings a 28.2$\times$ speedup over the conventional multi-way training method.code and data repo: \url{https://github.com/CONE-MT/Lego-MT.git}.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yuan-etal-2023-lego">
<titleInfo>
<title>Lego-MT: Learning Detachable Models for Massively Multilingual Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Fei</namePart>
<namePart type="family">Yuan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yinquan</namePart>
<namePart type="family">Lu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenhao</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lingpeng</namePart>
<namePart type="family">Kong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lei</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Qiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jingjing</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2023</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jordan</namePart>
<namePart type="family">Boyd-Graber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naoaki</namePart>
<namePart type="family">Okazaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Multilingual neural machine translation (MNMT) aims to build a unified model for many language directions. Existing monolithic models for MNMT encounter two challenges: parameter interference among languages and inefficient inference for large models. In this paper, we revisit the classic multi-way structures and develop a detachable model by assigning each language (or group of languages) to an individual branch that supports plug-and-play training and inference. To address the needs of learning representations for all languages in a unified space, we propose a novel efficient training recipe, upon which we build an effective detachable model, Lego-MT.For a fair comparison, we collect data from OPUS and build a translation benchmark covering 433 languages and 1.3B parallel data. Experiments show that Lego-MT with 1.2B parameters brings an average gain of 3.2 spBLEU. It even outperforms M2M-100 with 12B parameters. The proposed training recipe brings a 28.2\times speedup over the conventional multi-way training method.code and data repo: https://github.com/CONE-MT/Lego-MT.git.</abstract>
<identifier type="citekey">yuan-etal-2023-lego</identifier>
<identifier type="doi">10.18653/v1/2023.findings-acl.731</identifier>
<location>
<url>https://aclanthology.org/2023.findings-acl.731</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>11518</start>
<end>11533</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Lego-MT: Learning Detachable Models for Massively Multilingual Machine Translation
%A Yuan, Fei
%A Lu, Yinquan
%A Zhu, Wenhao
%A Kong, Lingpeng
%A Li, Lei
%A Qiao, Yu
%A Xu, Jingjing
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Findings of the Association for Computational Linguistics: ACL 2023
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F yuan-etal-2023-lego
%X Multilingual neural machine translation (MNMT) aims to build a unified model for many language directions. Existing monolithic models for MNMT encounter two challenges: parameter interference among languages and inefficient inference for large models. In this paper, we revisit the classic multi-way structures and develop a detachable model by assigning each language (or group of languages) to an individual branch that supports plug-and-play training and inference. To address the needs of learning representations for all languages in a unified space, we propose a novel efficient training recipe, upon which we build an effective detachable model, Lego-MT.For a fair comparison, we collect data from OPUS and build a translation benchmark covering 433 languages and 1.3B parallel data. Experiments show that Lego-MT with 1.2B parameters brings an average gain of 3.2 spBLEU. It even outperforms M2M-100 with 12B parameters. The proposed training recipe brings a 28.2\times speedup over the conventional multi-way training method.code and data repo: https://github.com/CONE-MT/Lego-MT.git.
%R 10.18653/v1/2023.findings-acl.731
%U https://aclanthology.org/2023.findings-acl.731
%U https://doi.org/10.18653/v1/2023.findings-acl.731
%P 11518-11533
Markdown (Informal)
[Lego-MT: Learning Detachable Models for Massively Multilingual Machine Translation](https://aclanthology.org/2023.findings-acl.731) (Yuan et al., Findings 2023)
ACL