@inproceedings{papacek-zabokrtsky-2026-tracing,
title = "Tracing Morph Origins in {C}zech: A Computational Approach to Morph-Level Etymology",
author = "Pap{\'a}{\v{c}}ek, Ale{\v{s}} Manuel Manuel and
{\v{Z}}abokrtsk{\'y}, Zden{\v{e}}k",
editor = "Sprugnoli, Rachele and
Passarotti, Marco",
booktitle = "Proceedings of the Fourth Workshop on Language Technologies for Historical and Ancient Languages ({LT}4{HALA} 2026) @ {LREC} 2026",
month = may,
year = "2026",
address = "Palma, Mallorca (Spain)",
publisher = "ELRA Language Resources Association (ELRA)",
url = "https://aclanthology.org/2026.lt4hala-1.2/",
doi = "10.63317/4h8ckod8kvq6",
pages = "7--18",
abstract = "Modern languages remain connected to ancient ones in multiple ways, including through etymology; for instance, Latin is among the most influential sources of borrowings in (modern) Czech, whether transmitted directly or mediated through other languages. This work focuses on predicting the etymological origin of individual morphs in Czech words. Given morphologically segmented Czech sentences, the task is to determine for each morph whether it is native or borrowed, and if borrowed, to identify the languages through which it entered Czech. Although some linguists have examined etymology at the level of individual morphs rather than whole words (Arkadiev et al., 2015), to our knowledge, no computational work has yet addressed this level of analysis. We created a manually annotated dataset of 300 Czech sentences comprising around 10,000 morphs with morph-level etymology labels, and trained supervised models using character-based and structural features. Our best lightweight system is a feed-forward neural network with a single hidden layer, trained on data augmented with entries from an etymological dictionary, reaching 96.2{\%} F1 on the test set. We also developed and tested several prompting variants for large language models; the best model Claude-Opus-4.5, achieved 97.8{\%} F1. We release the code, prompts, and dataset as open source at \url{https://github.com/ampapacek/MorphemeOrigin}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="papacek-zabokrtsky-2026-tracing">
<titleInfo>
<title>Tracing Morph Origins in Czech: A Computational Approach to Morph-Level Etymology</title>
</titleInfo>
<name type="personal">
<namePart type="given">Aleš</namePart>
<namePart type="given">Manuel</namePart>
<namePart type="given">Manuel</namePart>
<namePart type="family">Papáček</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zdeněk</namePart>
<namePart type="family">Žabokrtský</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 Fourth Workshop on Language Technologies for Historical and Ancient Languages (LT4HALA 2026) @ LREC 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rachele</namePart>
<namePart type="family">Sprugnoli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marco</namePart>
<namePart type="family">Passarotti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resources Association (ELRA)</publisher>
<place>
<placeTerm type="text">Palma, Mallorca (Spain)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Modern languages remain connected to ancient ones in multiple ways, including through etymology; for instance, Latin is among the most influential sources of borrowings in (modern) Czech, whether transmitted directly or mediated through other languages. This work focuses on predicting the etymological origin of individual morphs in Czech words. Given morphologically segmented Czech sentences, the task is to determine for each morph whether it is native or borrowed, and if borrowed, to identify the languages through which it entered Czech. Although some linguists have examined etymology at the level of individual morphs rather than whole words (Arkadiev et al., 2015), to our knowledge, no computational work has yet addressed this level of analysis. We created a manually annotated dataset of 300 Czech sentences comprising around 10,000 morphs with morph-level etymology labels, and trained supervised models using character-based and structural features. Our best lightweight system is a feed-forward neural network with a single hidden layer, trained on data augmented with entries from an etymological dictionary, reaching 96.2% F1 on the test set. We also developed and tested several prompting variants for large language models; the best model Claude-Opus-4.5, achieved 97.8% F1. We release the code, prompts, and dataset as open source at https://github.com/ampapacek/MorphemeOrigin.</abstract>
<identifier type="citekey">papacek-zabokrtsky-2026-tracing</identifier>
<identifier type="doi">10.63317/4h8ckod8kvq6</identifier>
<location>
<url>https://aclanthology.org/2026.lt4hala-1.2/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>7</start>
<end>18</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Tracing Morph Origins in Czech: A Computational Approach to Morph-Level Etymology
%A Papáček, Aleš Manuel Manuel
%A Žabokrtský, Zdeněk
%Y Sprugnoli, Rachele
%Y Passarotti, Marco
%S Proceedings of the Fourth Workshop on Language Technologies for Historical and Ancient Languages (LT4HALA 2026) @ LREC 2026
%D 2026
%8 May
%I ELRA Language Resources Association (ELRA)
%C Palma, Mallorca (Spain)
%F papacek-zabokrtsky-2026-tracing
%X Modern languages remain connected to ancient ones in multiple ways, including through etymology; for instance, Latin is among the most influential sources of borrowings in (modern) Czech, whether transmitted directly or mediated through other languages. This work focuses on predicting the etymological origin of individual morphs in Czech words. Given morphologically segmented Czech sentences, the task is to determine for each morph whether it is native or borrowed, and if borrowed, to identify the languages through which it entered Czech. Although some linguists have examined etymology at the level of individual morphs rather than whole words (Arkadiev et al., 2015), to our knowledge, no computational work has yet addressed this level of analysis. We created a manually annotated dataset of 300 Czech sentences comprising around 10,000 morphs with morph-level etymology labels, and trained supervised models using character-based and structural features. Our best lightweight system is a feed-forward neural network with a single hidden layer, trained on data augmented with entries from an etymological dictionary, reaching 96.2% F1 on the test set. We also developed and tested several prompting variants for large language models; the best model Claude-Opus-4.5, achieved 97.8% F1. We release the code, prompts, and dataset as open source at https://github.com/ampapacek/MorphemeOrigin.
%R 10.63317/4h8ckod8kvq6
%U https://aclanthology.org/2026.lt4hala-1.2/
%U https://doi.org/10.63317/4h8ckod8kvq6
%P 7-18
Markdown (Informal)
[Tracing Morph Origins in Czech: A Computational Approach to Morph-Level Etymology](https://aclanthology.org/2026.lt4hala-1.2/) (Papáček & Žabokrtský, LT4HALA 2026)
ACL