Elamur Research⌕
☰
home / papers / 2026 / From Lemmatization to Legal Terminology: Assessing an Hybrid Pipeline on Justinian's Digest

From Lemmatization to Legal Terminology: Assessing an Hybrid Pipeline on Justinian's Digest

Summary

The paper addresses lemmatization quality for Justinian's Digest, a large compilation of Roman legal writing. Accurate lemmas are needed because the authors plan to build a bilingual Latin–Italian legal glossary with Latin as the pivot language, as part of an ongoing Italian translation project. The Digest is hard for automatic tools. It has a dense specialised vocabulary, formulaic and non-classical syntax, passages in Ancient Greek, and textual gaps.

The authors use LatinPipe as the baseline and add GPT-4o-mini as a post-correction layer. They test this layer in three prompting setups. Outputs are checked against a manually annotated gold standard and against an expert-validated list of legal Latin terms.

All hybrid configurations beat the baseline, and few-shot prompting performs best. The gains are uneven. They concentrate on frequent lemmas and verb forms, which suggests the LLM mainly resolves morphologically ambiguous inflections.

The authors present the work as chiefly methodological. It shows how to evaluate lemmatization against downstream terminology goals rather than offering a universal fix for domain-specific lemmatization.

Key points

  • Corpus: 36 of the Digest's 50 books, about 635k tokens and roughly 37,000 sentences. The text follows the Mommsen–Krüger (1911) edition as supplied by the translators, and it was lowercased and stripped of diacritics before annotation.
  • Baseline: LatinPipe, an EvaLatin 2024 winner trained on several UD Latin treebanks (ITTB, LLCT, PROIEL, UDante, Perseus and others), reaches F1 = 95.05.
  • LLM settings: GPT-4o-mini was tested in three ways:
    • zero-shot without the LatinPipe lemma;
    • zero-shot with the LatinPipe lemma;
    • few-shot prompting, which gave the best results.
  • Evaluation resources: a gold standard of 3,703 manually annotated sentences and an expert-checked inventory of legal Latin technical terms.
  • Where it helps: improvements were significantly more likely for high-frequency lemmas and verbal forms. The authors caution that their experimental conditions may not carry over to real-world use.

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

Fields: Linguistics & languages · Natural language processing · Law · History & humanities

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/4x2mcetz9f5r

@inproceedings{marongiu2026lemmatization,
  title = {{From Lemmatization to Legal Terminology: Assessing an Hybrid Pipeline on Justinian's Digest}},
  author = {Marongiu, Paola and Sassolini, Eva},
  booktitle = {Proceedings of the Fourth Workshop on Language Technologies for Historical and Ancient Languages (LT4HALA 2026) @ LREC 2026},
  year = {2026},
  doi = {10.63317/4x2mcetz9f5r},
  url = {https://doi.org/10.63317/4x2mcetz9f5r}
}

References

36 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Thomas M Banchich, John Marenbon, and Charles J Reid Jr. 2015. The revival of roman law and canon law. In A Treatise of Legal Philosophy and General Jurisprudence: Volume 6: A History of the Philosophy of Law from the Ancient Greeks to the Scholastics, pages 251–265. Springer. find
  2. Marco Baroni. 2009. Distributions in text. In Anke Lüdeling and Merja Kytö, editors, Corpus linguistics: An international handbook. Volume 2, pages 803–821. Mouton de Gruyter, Berlin. find
  3. Ruirui Chen, Chengwei Qin, Weifeng Jiang, and Dongkyu Choi. 2024. Is a large language model a good annotator for event extraction? In Proceedings of the AAAI conference on artificial intelligence, volume 38, pages 17772–17780. find
  4. Frederick W Dingledy. 2016. The corpus juris civilis: a guide to its history and use. Legal Reference Services Quarterly, 35(4):231–255. find
  5. Margherita Fantoli, Marco Passarotti, Dominique Longrée, and Concepción Cabrillana. 2024. Lemmas in dialogue: Linking the lasla corpus to the lila knowledge base. Recent Trends and Findings in Latin Linguistics: Volume I: Syntax, Semantics and Pragmatics. Volume II: Semantics and Lexicography. Discourse and Dialogue, page 297. find
  6. Margherita Fantoli, Marco Passarotti, Francesco Mambrini, Giovanni Moretti, and Paolo Ruffolo. 2022. Linking the LASLA corpus in the LiLa knowledge base of interoperable linguistic resources for Latin. In Proceedings of the 8th Workshop on Linked Data in Linguistics within the 13th Language Resources and Evaluation Conference, pages 26–34, Marseille, France. European Language Resources Association. find
  7. Andrea Farina, Andrea Ballatore, and Barbara McGillivray. 2025. Mapping meaning in latin with large language models: A multi-task evaluation of preverbed motion verbs and spatial relation detection in llms. In CLiC-it 2025 Italian Conference on Computational Linguistics. Proceedings of the Eleventh Italian Conference on Computational Linguistics (CLiC-it 2025) Cagliari, Italy, September 24-26, 2025. find
  8. Eleonora Ghizzota, Paola Marongiu, Pierpaolo Basile, Stefano Ferilli, and Barbara McGillivray. 2026. Linguistic knowledge graphs for sense prediction: a case-study on latin. In Proceedings of the 15th Language Resources and Evaluation Conference. Anthology of Computers and the Humanities. find
  9. Saniya Irfan and Syed Juned Ali. 2025. Qallm: An llm-based ner dataset curation, annotation and evaluation in historical urdu elegies. In Computational Humanities Research 2025, pages 921– 936. Anthology of Computers and the Humanities. find
  10. Siyu Liang, Talant Mawkanuli, and Gina-Anne Levow. 2026. Hybrid neural-LLM pipeline for morphological glossing in endangered language documentation: A case study of jungar tuvan. In Proceedings of the Fifth Workshop on NLP Applications to Field Linguistics, pages 16–30, Rabat, Morocco. Association for Computational Linguistics. find
  11. Theodor Mommsen and P. Krüger, editors. 1911. Corpus iuris civilis: editio stereotypa duodecima. Institutiones, volume 1. Weidmannos, Berlin. find
  12. Kseniia Petukhova and Ekaterina Kochmar. 2025. A fully automated pipeline for conversational discourse annotation: Tree scheme generation and labeling with large language models. In Findings of the Association for Computational Linguistics: ACL 2025, pages 15829–15852. find
  13. Marton Ribary and Barbara McGillivray. 2020. A corpus approach to roman law based on justinian’s digest. In Informatics, volume 7, page 44. MDPI. find
  14. Frederick Riemenschneider. 2025. Beyond base predictors: Using LLMs to resolve ambiguities in Akkadian lemmatization. In Proceedings of the Second Workshop on Ancient Language Processing, pages 226–231, The Albuquerque Convention Center, Laguna. Association for Computational Linguistics. find our card
  15. Sandro Schipani. 2005. Iustiniani Augusti Digesta seu Pandectae. Digesti o Pandette dell’Imperatore Giustiniano. Testo e traduzione, volume 1. Giuffrè Editore. find
  16. Rachele Sprugnoli, Federica Iurescia, and Marco Passarotti. 2024. Overview of the EvaLatin 2024 evaluation campaign. In Proceedings of the Third Workshop on Language Technologies for Historical and Ancient Languages (LT4HALA) @ LREC-COLING-2024, pages 190–197, Torino, Italia. ELRA and ICCL. find
  17. Rachele Sprugnoli, Marco Passarotti, Flavio Massimiliano Cecchini, Margherita Fantoli, and Giovanni Moretti. 2022. Overview of the EvaLatin 2022 evaluation campaign. In Proceedings of the Second Workshop on Language Technologies for Historical and Ancient Languages, pages 183–188, Marseille, France. European Language Resources Association. find
  18. Rachele Sprugnoli, Marco Passarotti, Flavio Massimiliano Cecchini, and Matteo Pellegrini. 2020. Overview of the EvaLatin 2020 evaluation campaign. In Proceedings of LT4HALA 2020 - 1st Workshop on Language Technologies for Historical and Ancient Languages, pages 105– 110, Marseille, France. European Language Resources Association (ELRA). find
  19. Olia Toporkov, Alan Akbik, and Rodrigo Agerri. 2025. Lemma dilemma: On lemma generation without domain-or language-specific training data. arXiv preprint arXiv:2510.07434. ↗
  20. Martin Volk, Dominic Philipp Fischer, Lukas Fischer, Patricia Scheurer, and Phillip Benjamin Ströbel. 2024. LLM-based machine translation and summarization for Latin. In Proceedings of the Third Workshop on Language Technologies for Historical and Ancient Languages (LT4HALA) @ LREC-COLING-2024, pages 122–128, Torino, Italia. ELRA and ICCL. find
  21. Leonie Weissweiler, Abdullatif Köksal, and Hinrich Schütze. 2025. Hybrid human-llm corpus construction and llm evaluation for the caused-motion construction. Northern European Journal of Language Technology, 11(1):27–57. find
  22. George Kingsley Zipf. 1949. The Principle of Least Effort. Addison-Wesley. 11. find
  23. Language Resource References David Bamman and Gregory Crane. 2011. The ancient greek and latin dependency treebanks. In Language technology for cultural heritage: Selected papers from the LaTeCH Workshop Series, pages 79–98. Springer. find
  24. Flavio Massimiliano Cecchini, Timo Korkiakangas, and Marco Passarotti. 2020a. A new latin treebank for universal dependencies: Charters between ancient latin and romance languages. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 933–942. find
  25. Flavio Massimiliano Cecchini, Rachele Sprugnoli, Giovanni Moretti, and Marco Passarotti. 2020b. Udante: First steps towards the universal dependencies treebank of dante’s latin works. In Proceedings of the Seventh Italian Conference on Computational Linguistics (CLiC-it 2020), pages 66–72. find
  26. Margherita Fantoli and Miryam de Lhoneux. 2022. Linguistic annotation of neo-latin mathematical texts: a pilot-study to improve the automatic parsing of the archimedes latinus. In Proceedings of the Second Workshop on Language Technologies for Historical and Ancient Languages, pages 129–134. find
  27. Federica Gamba and Flavio Massimiliano Cecchini. 2024. De latinae linguae reparatione treebank. LINDAT/CLARIAH-CZ digital library at the Institute of Formal and Applied Linguistics (ÚFAL). find
  28. Federica Gamba and Daniel Zeman. 2023. Latin morphology through the centuries: Ensuring consistency for better language processing. In Proceedings of the Ancient Language Processing Workshop, pages 59–67. find
  29. Dag TT Haug and Marius Jøhndal. 2008. Creating a parallel treebank of the old indo-european bible translations. In Proceedings of the second workshop on language technology for cultural heritage data (LaTeCH 2008), pages 27– 34. Prague. find
  30. Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Jan Hajič, Christopher D. Manning, Sampo Pyysalo, Sebastian Schuster, Francis Tyers, and Daniel Zeman. 2020. Universal Dependencies v2: An evergrowing multilingual treebank collection. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 4034–4043, Marseille, France. European Language Resources Association. find our card
  31. Marco Passarotti. 2019. The project of the index thomisticus. Digital Classical Philology: Ancient Greek and Latin in the Digital Revolution, 10:299. find
  32. Marco Passarotti, Marco Budassi, Eleonora Litta, and Paolo Ruffolo. 2017. The lemlat 3.0 package for morphological analysis of Latin. In Proceedings of the NoDaLiDa 2017 Workshop on Processing Historical Language, pages 24– 31, Gothenburg. Linköping University Electronic Press. find
  33. Milan Straka. 2018. UDPipe 2.0 prototype at CoNLL 2018 UD shared task. In Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 197–207, Brussels, Belgium. Association for Computational Linguistics. find our card
  34. Milan Straka and Jana Straková. 2017. Tokenizing, POS tagging, lemmatizing and parsing UD 2.0 with UDPipe. In Proceedings of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 88–99, Vancouver, Canada. Association for Computational Linguistics. ↗ our card
  35. Milan Straka and Jana Straková. 2020. UD-Pipe at EvaLatin 2020: Contextualized embeddings and treebank embeddings. In Proceedings of LT4HALA 2020 - 1st Workshop on Language Technologies for Historical and Ancient Languages, pages 124–129, Marseille, France. European Language Resources Association (ELRA). find
  36. Milan Straka, Jana Straková, and Federica Gamba. 2024. ÚFAL LatinPipe at EvaLatin 2024: Morphosyntactic analysis of Latin. In Proceedings of the Third Workshop on Language Technologies for Historical and Ancient Languages (LT4HALA) @ LREC-COLING-2024, pages 207–214, Torino, Italia. ELRA and ICCL. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/marongiu-2026-lemmatization-legal-terminology-assessing-hybrid/