@inproceedings{khallaf-sharoff-2026-much,
title = "How Much Noise Can {BERT} Handle? Insights from Multilingual Sentence Difficulty Detection",
author = "Khallaf, Nouran and
Sharoff, Serge",
editor = "Piperidis, Stelios and
Bel, N{\'u}ria and
van den Heuvel, Henk and
Ide, Nancy and
Krek, Simon and
Toral, Antonio",
booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resource Association",
url = "https://aclanthology.org/2026.lrec-1.485/",
doi = "10.63317/3996hxsmfa2k",
pages = "6132--6143",
abstract = "Noisy training data can significantly degrade the performance of language-model-based classifiers, particularly in non-topical classification tasks. This study explores a range of denoising strategies for sentence-level difficulty detection, using training data derived from document-level difficulty annotations obtained through noisy crowdsourcing. Beyond monolingual settings, we also address cross-lingual transfer, where a multilingual language model is trained in one language and tested in another. We evaluate several noise reduction techniques, including Gaussian Mixture Models (GMM), Co-Teaching, Noise Transition Matrices, and Label Smoothing. Our results indicate that while BERT-based models exhibit inherent robustness to noise, incorporating explicit noise detection can further enhance performance. For our smaller dataset, GMM-based noise filtering proves particularly effective in improving prediction quality by raising the AUC score from 0.52 to 0.86, or to 0.92 when two de-noising methods are combined (GMM and Co-Teaching). However, for our larger dataset, the intrinsic regularisation of pre-trained language models provides a strong baseline, with denoising methods yielding only marginal gains (from 0.8948 to 0.8984, or to 0.9061 when two denoising methods are combined). Nonetheless, removing noisy sentences (about 20{\%} of the dataset) helps in producing a cleaner corpus with fewer infelicities. As a result we have released the largest available multilingual corpus for sentence difficulty prediction."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="khallaf-sharoff-2026-much">
<titleInfo>
<title>How Much Noise Can BERT Handle? Insights from Multilingual Sentence Difficulty Detection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nouran</namePart>
<namePart type="family">Khallaf</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Serge</namePart>
<namePart type="family">Sharoff</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fifteenth Language Resources and Evaluation Conference</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stelios</namePart>
<namePart type="family">Piperidis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Núria</namePart>
<namePart type="family">Bel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Henk</namePart>
<namePart type="family">van den Heuvel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nancy</namePart>
<namePart type="family">Ide</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Krek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Toral</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resource Association</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Noisy training data can significantly degrade the performance of language-model-based classifiers, particularly in non-topical classification tasks. This study explores a range of denoising strategies for sentence-level difficulty detection, using training data derived from document-level difficulty annotations obtained through noisy crowdsourcing. Beyond monolingual settings, we also address cross-lingual transfer, where a multilingual language model is trained in one language and tested in another. We evaluate several noise reduction techniques, including Gaussian Mixture Models (GMM), Co-Teaching, Noise Transition Matrices, and Label Smoothing. Our results indicate that while BERT-based models exhibit inherent robustness to noise, incorporating explicit noise detection can further enhance performance. For our smaller dataset, GMM-based noise filtering proves particularly effective in improving prediction quality by raising the AUC score from 0.52 to 0.86, or to 0.92 when two de-noising methods are combined (GMM and Co-Teaching). However, for our larger dataset, the intrinsic regularisation of pre-trained language models provides a strong baseline, with denoising methods yielding only marginal gains (from 0.8948 to 0.8984, or to 0.9061 when two denoising methods are combined). Nonetheless, removing noisy sentences (about 20% of the dataset) helps in producing a cleaner corpus with fewer infelicities. As a result we have released the largest available multilingual corpus for sentence difficulty prediction.</abstract>
<identifier type="citekey">khallaf-sharoff-2026-much</identifier>
<identifier type="doi">10.63317/3996hxsmfa2k</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.485/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>6132</start>
<end>6143</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T How Much Noise Can BERT Handle? Insights from Multilingual Sentence Difficulty Detection
%A Khallaf, Nouran
%A Sharoff, Serge
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F khallaf-sharoff-2026-much
%X Noisy training data can significantly degrade the performance of language-model-based classifiers, particularly in non-topical classification tasks. This study explores a range of denoising strategies for sentence-level difficulty detection, using training data derived from document-level difficulty annotations obtained through noisy crowdsourcing. Beyond monolingual settings, we also address cross-lingual transfer, where a multilingual language model is trained in one language and tested in another. We evaluate several noise reduction techniques, including Gaussian Mixture Models (GMM), Co-Teaching, Noise Transition Matrices, and Label Smoothing. Our results indicate that while BERT-based models exhibit inherent robustness to noise, incorporating explicit noise detection can further enhance performance. For our smaller dataset, GMM-based noise filtering proves particularly effective in improving prediction quality by raising the AUC score from 0.52 to 0.86, or to 0.92 when two de-noising methods are combined (GMM and Co-Teaching). However, for our larger dataset, the intrinsic regularisation of pre-trained language models provides a strong baseline, with denoising methods yielding only marginal gains (from 0.8948 to 0.8984, or to 0.9061 when two denoising methods are combined). Nonetheless, removing noisy sentences (about 20% of the dataset) helps in producing a cleaner corpus with fewer infelicities. As a result we have released the largest available multilingual corpus for sentence difficulty prediction.
%R 10.63317/3996hxsmfa2k
%U https://aclanthology.org/2026.lrec-1.485/
%U https://doi.org/10.63317/3996hxsmfa2k
%P 6132-6143
Markdown (Informal)
[How Much Noise Can BERT Handle? Insights from Multilingual Sentence Difficulty Detection](https://aclanthology.org/2026.lrec-1.485/) (Khallaf & Sharoff, LREC 2026)
ACL