@inproceedings{samuel-etal-2023-trained,
title = "Trained on 100 million words and still in shape: {BERT} meets {B}ritish {N}ational {C}orpus",
author = "Samuel, David and
Kutuzov, Andrey and
{\O}vrelid, Lilja and
Velldal, Erik",
editor = "Vlachos, Andreas and
Augenstein, Isabelle",
booktitle = "Findings of the Association for Computational Linguistics: EACL 2023",
month = may,
year = "2023",
address = "Dubrovnik, Croatia",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-eacl.146/",
doi = "10.18653/v1/2023.findings-eacl.146",
pages = "1954--1974",
abstract = "While modern masked language models (LMs) are trained on ever larger corpora, we here explore the effects of down-scaling training to a modestly-sized but representative, well-balanced, and publicly available English text source {--} the British National Corpus. We show that pre-training on this carefully curated corpus can reach better performance than the original BERT model. We argue that this type of corpora has great potential as a language modeling benchmark. To showcase this potential, we present fair, reproducible and data-efficient comparative studies of LMs, in which we evaluate several training objectives and model architectures and replicate previous empirical results in a systematic way. We propose an optimized LM architecture called LTG-BERT."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="samuel-etal-2023-trained">
<titleInfo>
<title>Trained on 100 million words and still in shape: BERT meets British National Corpus</title>
</titleInfo>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Samuel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrey</namePart>
<namePart type="family">Kutuzov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lilja</namePart>
<namePart type="family">Øvrelid</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Erik</namePart>
<namePart type="family">Velldal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EACL 2023</title>
</titleInfo>
<name type="personal">
<namePart type="given">Andreas</namePart>
<namePart type="family">Vlachos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Isabelle</namePart>
<namePart type="family">Augenstein</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Dubrovnik, Croatia</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>While modern masked language models (LMs) are trained on ever larger corpora, we here explore the effects of down-scaling training to a modestly-sized but representative, well-balanced, and publicly available English text source – the British National Corpus. We show that pre-training on this carefully curated corpus can reach better performance than the original BERT model. We argue that this type of corpora has great potential as a language modeling benchmark. To showcase this potential, we present fair, reproducible and data-efficient comparative studies of LMs, in which we evaluate several training objectives and model architectures and replicate previous empirical results in a systematic way. We propose an optimized LM architecture called LTG-BERT.</abstract>
<identifier type="citekey">samuel-etal-2023-trained</identifier>
<identifier type="doi">10.18653/v1/2023.findings-eacl.146</identifier>
<location>
<url>https://aclanthology.org/2023.findings-eacl.146/</url>
</location>
<part>
<date>2023-05</date>
<extent unit="page">
<start>1954</start>
<end>1974</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Trained on 100 million words and still in shape: BERT meets British National Corpus
%A Samuel, David
%A Kutuzov, Andrey
%A Øvrelid, Lilja
%A Velldal, Erik
%Y Vlachos, Andreas
%Y Augenstein, Isabelle
%S Findings of the Association for Computational Linguistics: EACL 2023
%D 2023
%8 May
%I Association for Computational Linguistics
%C Dubrovnik, Croatia
%F samuel-etal-2023-trained
%X While modern masked language models (LMs) are trained on ever larger corpora, we here explore the effects of down-scaling training to a modestly-sized but representative, well-balanced, and publicly available English text source – the British National Corpus. We show that pre-training on this carefully curated corpus can reach better performance than the original BERT model. We argue that this type of corpora has great potential as a language modeling benchmark. To showcase this potential, we present fair, reproducible and data-efficient comparative studies of LMs, in which we evaluate several training objectives and model architectures and replicate previous empirical results in a systematic way. We propose an optimized LM architecture called LTG-BERT.
%R 10.18653/v1/2023.findings-eacl.146
%U https://aclanthology.org/2023.findings-eacl.146/
%U https://doi.org/10.18653/v1/2023.findings-eacl.146
%P 1954-1974
Markdown (Informal)
[Trained on 100 million words and still in shape: BERT meets British National Corpus](https://aclanthology.org/2023.findings-eacl.146/) (Samuel et al., Findings 2023)
ACL