Elamur Research⌕
☰
home / papers / 2026 / Tracing Morph Origins in Czech: A Computational Approach to Morph-Level Etymology

Tracing Morph Origins in Czech: A Computational Approach to Morph-Level Etymology

Paper · 2026 · Aleš Manuel Manuel Papáček and Zdeněk Žabokrtský

Summary

Etymological resources usually describe whole words. Many words, however, are hybrids that mix native and borrowed parts, such as a foreign root with a native ending. This paper is presented as the first computational treatment of etymology at the morph level, applied to Czech. Given a morphologically segmented sentence, each morph is labelled as native or borrowed. For borrowed morphs, the system also gives the chain of languages through which the morph reached Czech.

The authors manually annotated 300 sentences from the SIGMORPHON 2022 segmentation data. They compared simple baselines and dictionary rules based on the Czech Etymological Lexicon (CzEtyL) with classifiers using character n-grams, morph type and position in the word. They also tested several LLMs using prompts only.

A small MLP trained with extra dictionary-derived data reached 96.2 % F1, and Claude-Opus-4.5 reached 97.8 %. The work offers a fine-grained, reusable way to study borrowing and language contact. In principle it could be applied to other languages that have morph-level annotations.

Key points

  • Dataset: about 10,000 morphs in 300 Czech sentences (SIGMORPHON 2022), each labelled as native or with its full borrowing path. Example: architekt came from Greek via Latin and German.
  • Resources used: CzEtyL, the digital version of Rejzek's Czech Etymological Dictionary with over 11,000 main entries, extended through derivational links from DeriNet.
  • Methods compared: majority-label, memorisation and rule-based baselines; LR, SVM and MLP classifiers; embeddings; semi-supervised self-training; prompting of LLMs.
  • Best results: an MLP with one hidden layer of 30 units achieved 96.2 % F1 on the test set; Claude-Opus-4.5 achieved 97.8 % F1. Code, prompts and data are open at github.com/ampapacek/MorphemeOrigin.
  • Case study: the MLP was applied to an exploratory Czech legal corpus from e-Sbírka covering 1918–2026 (109,257 words).
    • On average, 88.1 % of morphs were predicted native and 8.96 % borrowed; Latin was the leading source at about 5.6 %.
    • The borrowed share rises slightly over time, by roughly 2.5 percentage points per century.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Charles UniversityPrague · Czechia133 works · 20,686 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 1Research subjects and languages the work is about.Historical linguistics & language change · 249 works

Cite

Cite as Aleš Manuel Manuel Papáček and Zdeněk Žabokrtský. 2026. Tracing Morph Origins in Czech: A Computational Approach to Morph-Level Etymology. Proceedings of the Fourth Workshop on Language Technologies for Historical and Ancient Languages (LT4HALA 2026) @ LREC 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/4h8ckod8kvq6

@inproceedings{papacek2026tracing,
  title = {{Tracing Morph Origins in Czech: A Computational Approach to Morph-Level Etymology}},
  author = {Aleš Manuel Manuel Papáček and Zdeněk Žabokrtský},
  booktitle = {Proceedings of the Fourth Workshop on Language Technologies for Historical and Ancient Languages (LT4HALA 2026) @ LREC 2026},
  year = {2026},
  doi = {10.63317/4h8ckod8kvq6},
  url = {https://doi.org/10.63317/4h8ckod8kvq6}
}

References

9 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Vojtěch John. 2024. Morph classifier. Master’s thesis, Charles University, Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics (ÚFAL), Prague, Czech Republic. find
  2. Peter Arkadiev, Francesco Gardani, and Nino Amiridze. 2015. Borrowed Morphology: An Overview, pages 1–23. De Gruyter Mouton. find
  3. Viktória Kovácsová. 2025. Automatic detection of lexical borrowings. Master’s thesis, Charles University, Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics (ÚFAL). find
  4. Mark Aronoff and Kirsten Fudeman. 2011. What is Morphology? Fundamentals of Linguistics. Wiley. find
  5. Oxford English Dictionary. 2025. school, n. https://www.oed.com/dictionary/school_n1. Oxford University Press, accessed 13 Oct 2025. ↗
  6. Jiří Rejzek. 2019. Český etymologický slovník. 3rd edition. LEDA. find
  7. Frank Seifart. 2015. Direct and indirect affix borrowing. Language, 91(3):511–532. find
  8. Wiktionary. 2025. Wiktionary: The free dictionary. https://www.wiktionary.org/. Accessed: 2025-10-04. ↗
  9. A. Etymological Analysis of a Diachronic Czech Legal Corpus find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/papacek-2026-tracing-morph-origins-czech-computational/