@inproceedings{morozov-etal-2026-belamorphlm,
title = "{B}ela{M}orph{LM}: A Controlled Study of Morphology-Aware Tokenization for {B}elarusian Encoder Language Models",
author = "Morozov, Dmitry and
Astapenka, Lizaveta and
Shcherbakova, Olga",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.708/",
doi = "10.18653/2026.findings-emnlp.708",
pages = "14867--14885",
ISBN = "979-8-89176-462-0",
abstract = "Tokenization is a central component of pretrained language models, especially for morphologically rich low-resource languages. We present a controlled study of morphology-aware tokenization for Belarusian masked language models. To isolate the effect of tokenization, we pretrain three RoBERTa-family models on the same Belarusian HPLT corpus, varying only the tokenizer: standard WordPiece, WordPiece trained on morpheme-segmented text, and a dictionary-based hybrid morpheme tokenizer with WordPiece fallback. We evaluate the models on masked language modeling, Universal Dependencies morphology-sensitive tasks, and BelarusianGLUE. Morphology-aware tokenization substantially improves morpheme-boundary preservation and yields clear gains on explicitly morphological tasks, especially fine-grained feature prediction and lemmatization, where models must recover information encoded in inflectional structure. However, these gains do not uniformly transfer to broad sentence-level NLU: on BelarusianGLUE, morphology-aware tokenizers improve some tasks but remain comparable to or weaker than standard WordPiece on others. These findings suggest that morpheme-informed tokenization is a targeted inductive bias rather than a universal replacement for statistical subword segmentation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="morozov-etal-2026-belamorphlm">
<titleInfo>
<title>BelaMorphLM: A Controlled Study of Morphology-Aware Tokenization for Belarusian Encoder Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Dmitry</namePart>
<namePart type="family">Morozov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lizaveta</namePart>
<namePart type="family">Astapenka</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Olga</namePart>
<namePart type="family">Shcherbakova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Tokenization is a central component of pretrained language models, especially for morphologically rich low-resource languages. We present a controlled study of morphology-aware tokenization for Belarusian masked language models. To isolate the effect of tokenization, we pretrain three RoBERTa-family models on the same Belarusian HPLT corpus, varying only the tokenizer: standard WordPiece, WordPiece trained on morpheme-segmented text, and a dictionary-based hybrid morpheme tokenizer with WordPiece fallback. We evaluate the models on masked language modeling, Universal Dependencies morphology-sensitive tasks, and BelarusianGLUE. Morphology-aware tokenization substantially improves morpheme-boundary preservation and yields clear gains on explicitly morphological tasks, especially fine-grained feature prediction and lemmatization, where models must recover information encoded in inflectional structure. However, these gains do not uniformly transfer to broad sentence-level NLU: on BelarusianGLUE, morphology-aware tokenizers improve some tasks but remain comparable to or weaker than standard WordPiece on others. These findings suggest that morpheme-informed tokenization is a targeted inductive bias rather than a universal replacement for statistical subword segmentation.</abstract>
<identifier type="citekey">morozov-etal-2026-belamorphlm</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.708</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.708/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>14867</start>
<end>14885</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T BelaMorphLM: A Controlled Study of Morphology-Aware Tokenization for Belarusian Encoder Language Models
%A Morozov, Dmitry
%A Astapenka, Lizaveta
%A Shcherbakova, Olga
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F morozov-etal-2026-belamorphlm
%X Tokenization is a central component of pretrained language models, especially for morphologically rich low-resource languages. We present a controlled study of morphology-aware tokenization for Belarusian masked language models. To isolate the effect of tokenization, we pretrain three RoBERTa-family models on the same Belarusian HPLT corpus, varying only the tokenizer: standard WordPiece, WordPiece trained on morpheme-segmented text, and a dictionary-based hybrid morpheme tokenizer with WordPiece fallback. We evaluate the models on masked language modeling, Universal Dependencies morphology-sensitive tasks, and BelarusianGLUE. Morphology-aware tokenization substantially improves morpheme-boundary preservation and yields clear gains on explicitly morphological tasks, especially fine-grained feature prediction and lemmatization, where models must recover information encoded in inflectional structure. However, these gains do not uniformly transfer to broad sentence-level NLU: on BelarusianGLUE, morphology-aware tokenizers improve some tasks but remain comparable to or weaker than standard WordPiece on others. These findings suggest that morpheme-informed tokenization is a targeted inductive bias rather than a universal replacement for statistical subword segmentation.
%R 10.18653/2026.findings-emnlp.708
%U https://aclanthology.org/2026.findings-emnlp.708/
%U https://doi.org/10.18653/2026.findings-emnlp.708
%P 14867-14885
Markdown (Informal)
[BelaMorphLM: A Controlled Study of Morphology-Aware Tokenization for Belarusian Encoder Language Models](https://aclanthology.org/2026.findings-emnlp.708/) (Morozov et al., Findings 2026)
ACL