@inproceedings{wolleb-etal-2023-assessing,
title = "Assessing the Importance of Frequency versus Compositionality for Subword-based Tokenization in {NMT}",
author = "Wolleb, Benoist and
Silvestri, Romain and
Vernikos, Georgios and
Dolamic, Ljiljana and
Popescu-Belis, Andrei",
editor = "Nurminen, Mary and
Brenner, Judith and
Koponen, Maarit and
Latomaa, Sirkku and
Mikhailov, Mikhail and
Schierl, Frederike and
Ranasinghe, Tharindu and
Vanmassenhove, Eva and
Vidal, Sergi Alvarez and
Aranberri, Nora and
Nunziatini, Mara and
Escart{\'\i}n, Carla Parra and
Forcada, Mikel and
Popovic, Maja and
Scarton, Carolina and
Moniz, Helena",
booktitle = "Proceedings of the 24th Annual Conference of the European Association for Machine Translation",
month = jun,
year = "2023",
address = "Tampere, Finland",
publisher = "European Association for Machine Translation",
url = "https://aclanthology.org/2023.eamt-1.14",
pages = "137--146",
abstract = "Subword tokenization is the de-facto standard for tokenization in neural language models and machine translation systems. Three advantages are frequently put forward in favor of subwords: shorter encoding of frequent tokens, compositionality of subwords, and ability to deal with unknown words. As their relative importance is not entirely clear yet, we propose a tokenization approach that enables us to separate frequency (the first advantage) from compositionality, thanks to the use of Huffman coding, which tokenizes words using a fixed amount of symbols. Experiments with CS-DE, EN-FR and EN-DE NMT show that frequency alone accounts for approximately 90{\%} of the BLEU scores reached by BPE, hence compositionality has less importance than previously thought.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wolleb-etal-2023-assessing">
<titleInfo>
<title>Assessing the Importance of Frequency versus Compositionality for Subword-based Tokenization in NMT</title>
</titleInfo>
<name type="personal">
<namePart type="given">Benoist</namePart>
<namePart type="family">Wolleb</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Romain</namePart>
<namePart type="family">Silvestri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Georgios</namePart>
<namePart type="family">Vernikos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ljiljana</namePart>
<namePart type="family">Dolamic</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrei</namePart>
<namePart type="family">Popescu-Belis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-06</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 24th Annual Conference of the European Association for Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mary</namePart>
<namePart type="family">Nurminen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Judith</namePart>
<namePart type="family">Brenner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maarit</namePart>
<namePart type="family">Koponen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sirkku</namePart>
<namePart type="family">Latomaa</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mikhail</namePart>
<namePart type="family">Mikhailov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Frederike</namePart>
<namePart type="family">Schierl</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tharindu</namePart>
<namePart type="family">Ranasinghe</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Eva</namePart>
<namePart type="family">Vanmassenhove</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sergi</namePart>
<namePart type="given">Alvarez</namePart>
<namePart type="family">Vidal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nora</namePart>
<namePart type="family">Aranberri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mara</namePart>
<namePart type="family">Nunziatini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carla</namePart>
<namePart type="given">Parra</namePart>
<namePart type="family">Escartín</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mikel</namePart>
<namePart type="family">Forcada</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maja</namePart>
<namePart type="family">Popovic</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carolina</namePart>
<namePart type="family">Scarton</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Helena</namePart>
<namePart type="family">Moniz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>European Association for Machine Translation</publisher>
<place>
<placeTerm type="text">Tampere, Finland</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Subword tokenization is the de-facto standard for tokenization in neural language models and machine translation systems. Three advantages are frequently put forward in favor of subwords: shorter encoding of frequent tokens, compositionality of subwords, and ability to deal with unknown words. As their relative importance is not entirely clear yet, we propose a tokenization approach that enables us to separate frequency (the first advantage) from compositionality, thanks to the use of Huffman coding, which tokenizes words using a fixed amount of symbols. Experiments with CS-DE, EN-FR and EN-DE NMT show that frequency alone accounts for approximately 90% of the BLEU scores reached by BPE, hence compositionality has less importance than previously thought.</abstract>
<identifier type="citekey">wolleb-etal-2023-assessing</identifier>
<location>
<url>https://aclanthology.org/2023.eamt-1.14</url>
</location>
<part>
<date>2023-06</date>
<extent unit="page">
<start>137</start>
<end>146</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Assessing the Importance of Frequency versus Compositionality for Subword-based Tokenization in NMT
%A Wolleb, Benoist
%A Silvestri, Romain
%A Vernikos, Georgios
%A Dolamic, Ljiljana
%A Popescu-Belis, Andrei
%Y Nurminen, Mary
%Y Brenner, Judith
%Y Koponen, Maarit
%Y Latomaa, Sirkku
%Y Mikhailov, Mikhail
%Y Schierl, Frederike
%Y Ranasinghe, Tharindu
%Y Vanmassenhove, Eva
%Y Vidal, Sergi Alvarez
%Y Aranberri, Nora
%Y Nunziatini, Mara
%Y Escartín, Carla Parra
%Y Forcada, Mikel
%Y Popovic, Maja
%Y Scarton, Carolina
%Y Moniz, Helena
%S Proceedings of the 24th Annual Conference of the European Association for Machine Translation
%D 2023
%8 June
%I European Association for Machine Translation
%C Tampere, Finland
%F wolleb-etal-2023-assessing
%X Subword tokenization is the de-facto standard for tokenization in neural language models and machine translation systems. Three advantages are frequently put forward in favor of subwords: shorter encoding of frequent tokens, compositionality of subwords, and ability to deal with unknown words. As their relative importance is not entirely clear yet, we propose a tokenization approach that enables us to separate frequency (the first advantage) from compositionality, thanks to the use of Huffman coding, which tokenizes words using a fixed amount of symbols. Experiments with CS-DE, EN-FR and EN-DE NMT show that frequency alone accounts for approximately 90% of the BLEU scores reached by BPE, hence compositionality has less importance than previously thought.
%U https://aclanthology.org/2023.eamt-1.14
%P 137-146
Markdown (Informal)
[Assessing the Importance of Frequency versus Compositionality for Subword-based Tokenization in NMT](https://aclanthology.org/2023.eamt-1.14) (Wolleb et al., EAMT 2023)
ACL