@inproceedings{zhang-etal-2022-robust,
title = "How Robust is Neural Machine Translation to Language Imbalance in Multilingual Tokenizer Training?",
author = "Zhang, Shiyue and
Chaudhary, Vishrav and
Goyal, Naman and
Cross, James and
Wenzek, Guillaume and
Bansal, Mohit and
Guzman, Francisco",
editor = "Duh, Kevin and
Guzm{\'a}n, Francisco",
booktitle = "Proceedings of the 15th biennial conference of the Association for Machine Translation in the Americas (Volume 1: Research Track)",
month = sep,
year = "2022",
address = "Orlando, USA",
publisher = "Association for Machine Translation in the Americas",
url = "https://aclanthology.org/2022.amta-research.8",
pages = "97--116",
abstract = "A multilingual tokenizer is a fundamental component of multilingual neural machine translation. It is trained from a multilingual corpus. Since a skewed data distribution is considered to be harmful, a sampling strategy is usually used to balance languages in the corpus. However, few works have systematically answered how language imbalance in tokenizer training affects downstream performance. In this work, we analyze how translation performance changes as the data ratios among languages vary in the tokenizer training corpus. We find that while relatively better performance is often observed when languages are more equally sampled, the downstream performance is more robust to language imbalance than we usually expected. Two features, UNK rate and closeness to the character level, can warn of poor downstream performance before performing the task. We also distinguish language sampling for tokenizer training from sampling for model training and show that the model is more sensitive to the latter.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2022-robust">
<titleInfo>
<title>How Robust is Neural Machine Translation to Language Imbalance in Multilingual Tokenizer Training?</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shiyue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vishrav</namePart>
<namePart type="family">Chaudhary</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naman</namePart>
<namePart type="family">Goyal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">James</namePart>
<namePart type="family">Cross</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guillaume</namePart>
<namePart type="family">Wenzek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Francisco</namePart>
<namePart type="family">Guzman</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2022-09</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th biennial conference of the Association for Machine Translation in the Americas (Volume 1: Research Track)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kevin</namePart>
<namePart type="family">Duh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Francisco</namePart>
<namePart type="family">Guzmán</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Machine Translation in the Americas</publisher>
<place>
<placeTerm type="text">Orlando, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>A multilingual tokenizer is a fundamental component of multilingual neural machine translation. It is trained from a multilingual corpus. Since a skewed data distribution is considered to be harmful, a sampling strategy is usually used to balance languages in the corpus. However, few works have systematically answered how language imbalance in tokenizer training affects downstream performance. In this work, we analyze how translation performance changes as the data ratios among languages vary in the tokenizer training corpus. We find that while relatively better performance is often observed when languages are more equally sampled, the downstream performance is more robust to language imbalance than we usually expected. Two features, UNK rate and closeness to the character level, can warn of poor downstream performance before performing the task. We also distinguish language sampling for tokenizer training from sampling for model training and show that the model is more sensitive to the latter.</abstract>
<identifier type="citekey">zhang-etal-2022-robust</identifier>
<location>
<url>https://aclanthology.org/2022.amta-research.8</url>
</location>
<part>
<date>2022-09</date>
<extent unit="page">
<start>97</start>
<end>116</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T How Robust is Neural Machine Translation to Language Imbalance in Multilingual Tokenizer Training?
%A Zhang, Shiyue
%A Chaudhary, Vishrav
%A Goyal, Naman
%A Cross, James
%A Wenzek, Guillaume
%A Bansal, Mohit
%A Guzman, Francisco
%Y Duh, Kevin
%Y Guzmán, Francisco
%S Proceedings of the 15th biennial conference of the Association for Machine Translation in the Americas (Volume 1: Research Track)
%D 2022
%8 September
%I Association for Machine Translation in the Americas
%C Orlando, USA
%F zhang-etal-2022-robust
%X A multilingual tokenizer is a fundamental component of multilingual neural machine translation. It is trained from a multilingual corpus. Since a skewed data distribution is considered to be harmful, a sampling strategy is usually used to balance languages in the corpus. However, few works have systematically answered how language imbalance in tokenizer training affects downstream performance. In this work, we analyze how translation performance changes as the data ratios among languages vary in the tokenizer training corpus. We find that while relatively better performance is often observed when languages are more equally sampled, the downstream performance is more robust to language imbalance than we usually expected. Two features, UNK rate and closeness to the character level, can warn of poor downstream performance before performing the task. We also distinguish language sampling for tokenizer training from sampling for model training and show that the model is more sensitive to the latter.
%U https://aclanthology.org/2022.amta-research.8
%P 97-116
Markdown (Informal)
[How Robust is Neural Machine Translation to Language Imbalance in Multilingual Tokenizer Training?](https://aclanthology.org/2022.amta-research.8) (Zhang et al., AMTA 2022)
ACL