@inproceedings{h-de-nardi-2026-multilingual,
title = "When Multilingual Evaluation Assumptions Fail: Tokenization Effects Across Scripts",
author = "H, Manodyna K and
De Nardi, Luc",
editor = "Hettiarachchi, Hansi and
Ranasinghe, Tharindu and
Plum, Alistair and
Rayson, Paul and
Mitkov, Ruslan and
Gaber, Mohamed and
Premasiri, Damith and
Tan, Fiona Anting and
Uyangodage, Lasitha",
booktitle = "Proceedings of the Second Workshop on Language Models for Low-Resource Languages ({L}o{R}es{LM} 2026)",
month = mar,
year = "2026",
address = "Rabat, Morocco",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.loreslm-1.4/",
pages = "41--49",
ISBN = "979-8-89176-377-7",
abstract = "Multilingual evaluation often relies on language coverage or translated benchmarks, implicitly assuming that subword tokenization behaves comparably across scripts. In mixed-script settings, this assumption breaks down. We examine this effect using polarity detection as a case study, comparing Orthographic Syllable Pair Encoding (OSPE) and Byte Pair Encoding (BPE) under identical architectures, data, and training conditions on SemEval Task 9, which spans Devanagari, Perso-Arabic, and Latin scripts. OSPE is applied to Hindi, Nepali, Urdu, and Arabic, while BPE is retained for English. We find that BPE systematically underestimates performance in abugida and abjad scripts, producing fragmented representations, unstable optimization, and drops of up to 27 macro-F1 points for Nepali, while English remains largely unaffected. Script-aware segmentation preserves orthographic structure, stabilizes training, and improves cross-language comparability without additional data or model scaling, highlighting tokenization as a latent but consequential evaluation decision in multilingual benchmarks."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="h-de-nardi-2026-multilingual">
<titleInfo>
<title>When Multilingual Evaluation Assumptions Fail: Tokenization Effects Across Scripts</title>
</titleInfo>
<name type="personal">
<namePart type="given">Manodyna</namePart>
<namePart type="given">K</namePart>
<namePart type="family">H</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Luc</namePart>
<namePart type="family">De Nardi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-03</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Second Workshop on Language Models for Low-Resource Languages (LoResLM 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hansi</namePart>
<namePart type="family">Hettiarachchi</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">Alistair</namePart>
<namePart type="family">Plum</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paul</namePart>
<namePart type="family">Rayson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruslan</namePart>
<namePart type="family">Mitkov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohamed</namePart>
<namePart type="family">Gaber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Damith</namePart>
<namePart type="family">Premasiri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fiona</namePart>
<namePart type="given">Anting</namePart>
<namePart type="family">Tan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lasitha</namePart>
<namePart type="family">Uyangodage</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Rabat, Morocco</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-377-7</identifier>
</relatedItem>
<abstract>Multilingual evaluation often relies on language coverage or translated benchmarks, implicitly assuming that subword tokenization behaves comparably across scripts. In mixed-script settings, this assumption breaks down. We examine this effect using polarity detection as a case study, comparing Orthographic Syllable Pair Encoding (OSPE) and Byte Pair Encoding (BPE) under identical architectures, data, and training conditions on SemEval Task 9, which spans Devanagari, Perso-Arabic, and Latin scripts. OSPE is applied to Hindi, Nepali, Urdu, and Arabic, while BPE is retained for English. We find that BPE systematically underestimates performance in abugida and abjad scripts, producing fragmented representations, unstable optimization, and drops of up to 27 macro-F1 points for Nepali, while English remains largely unaffected. Script-aware segmentation preserves orthographic structure, stabilizes training, and improves cross-language comparability without additional data or model scaling, highlighting tokenization as a latent but consequential evaluation decision in multilingual benchmarks.</abstract>
<identifier type="citekey">h-de-nardi-2026-multilingual</identifier>
<location>
<url>https://aclanthology.org/2026.loreslm-1.4/</url>
</location>
<part>
<date>2026-03</date>
<extent unit="page">
<start>41</start>
<end>49</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T When Multilingual Evaluation Assumptions Fail: Tokenization Effects Across Scripts
%A H, Manodyna K.
%A De Nardi, Luc
%Y Hettiarachchi, Hansi
%Y Ranasinghe, Tharindu
%Y Plum, Alistair
%Y Rayson, Paul
%Y Mitkov, Ruslan
%Y Gaber, Mohamed
%Y Premasiri, Damith
%Y Tan, Fiona Anting
%Y Uyangodage, Lasitha
%S Proceedings of the Second Workshop on Language Models for Low-Resource Languages (LoResLM 2026)
%D 2026
%8 March
%I Association for Computational Linguistics
%C Rabat, Morocco
%@ 979-8-89176-377-7
%F h-de-nardi-2026-multilingual
%X Multilingual evaluation often relies on language coverage or translated benchmarks, implicitly assuming that subword tokenization behaves comparably across scripts. In mixed-script settings, this assumption breaks down. We examine this effect using polarity detection as a case study, comparing Orthographic Syllable Pair Encoding (OSPE) and Byte Pair Encoding (BPE) under identical architectures, data, and training conditions on SemEval Task 9, which spans Devanagari, Perso-Arabic, and Latin scripts. OSPE is applied to Hindi, Nepali, Urdu, and Arabic, while BPE is retained for English. We find that BPE systematically underestimates performance in abugida and abjad scripts, producing fragmented representations, unstable optimization, and drops of up to 27 macro-F1 points for Nepali, while English remains largely unaffected. Script-aware segmentation preserves orthographic structure, stabilizes training, and improves cross-language comparability without additional data or model scaling, highlighting tokenization as a latent but consequential evaluation decision in multilingual benchmarks.
%U https://aclanthology.org/2026.loreslm-1.4/
%P 41-49
Markdown (Informal)
[When Multilingual Evaluation Assumptions Fail: Tokenization Effects Across Scripts](https://aclanthology.org/2026.loreslm-1.4/) (H & De Nardi, LoResLM 2026)
ACL