Simulating the Evolution of Grammatical Gender from Latin to Old Occitan: A Computational Approach Using LSTM with Attention
Authors
Marinus Wiedner, Matthias SchöffelKey Information
Journal
Published in Vol. 3 No. 1 (2026) of AI-Linguistica. Linguistic Studies on AI-Generated Texts and Discourses.
AI-Linguistica. Linguistic Studies on AI-Generated Texts and Discourses is a new scholarly journal aiming at providing a publishing plateform for researchers from all areas of Linguistics (interfacing with neighboring fields: Communication Science, Media and Journalism Studies, Computational Linguistics) to reflect on generated texts from a variety of perspectives: theoretical, descriptive, and applied.
We understand ‘generated texts’ in a broad sense, including formats as diverse as texts generated by Large Language Models, AI-powered smart agents (i.e. chatbots, voice assistants, social bots etc.), writing assistance tools, template-based software, and neural machine translation services.
About the JournalAbstract
In this article, we present a first approach on simulating the change of gender from Latin to Old Occitan. The reduction of genders in the transition to the Romance languages is one of the most important developments in the nominal system. In order to simulate this language change, we used varied data, taken from the most exhaustive Old Occitan dictionary (DOM), from an edition of a juridical text as well as from the transcriptions of two manuscripts taken from COMETA (Wiedner 2025). The nouns with information on their etymon and the gender in Latin and Old Occitan were then used as input for the simulation. Building on previous simulation-based approaches (Polinsky and Van Everbroeck 2003), our model provides a form-based, data-driven perspective on how gender marking might evolve under the interaction of morphological cues. Despite its inherent simplifications, the model achieved good predictive accuracy and yielded interpretable tendencies in the distribution of attention, suggesting that even at the level of orthographic form, systematic regularities relevant to gender can be captured computationally.
References
Allassonnière-Tang, Marc & Basirat, Ali. 2020. Word embedding and neural network on grammatical gender – A case study of Swedish. arXiv:2007.14222. https://doi.org/10.48550/arXiv.2007.14222
Allassonnière-Tang, Marc & Brown, Dunstan & Fedden, Sebastian. 2021. Testing Semantic Dominance in Mian Gender: Three Machine Learning Models. Oceanic Linguistics 60 (2), 302–334. https://hal.science/hal-03509042v1/document DOI: https://doi.org/10.1353/ol.2021.0018
Blom, Alderik H. 2023. Gaulish in the Late Empire (c. 200–600 CE). In Mullen, Alex & Woudhuysen, George (eds.): Languages and Communities in the Late-Roman and Post-Imperial Western Provinces. Oxford: Oxford University Press, 129–154. DOI: https://doi.org/10.1093/oso/9780198888956.003.0005
Burns, Patrick J. 2023. LatinCy: Synthetic Trained Pipelines for Latin NLP. ArXiv abs/2305.04365. https://doi.org/10.48550/arXiv.2305.04365
Chambon, Jean-Pierre. 2003. La déclinaison en ancien occitan, ou: comment s’en débarrasser ? : Une réanalyse descriptive non orthodoxe de la flexion substantivale. Revue de Linguistique Romane 67 (267–268): 343–364.
Chircu-Buftea, Adrian. 2011. Précis de morphologie romane. Cluj-Napoca: Casa Cartii de Stiinta.
Coker, Amy. 2009. Analogical change and grammatical gender in ancient Greek. Journal of Greek Linguistics 9 (1), 34–55. https://www.sciencedirect.com/org/science/article/pii/S1566584409000026 DOI: https://doi.org/10.1163/156658409X12529372103263
COMETA = Wiedner, Marinus (ed.). 2025. Corpus de l’occitan médiéval comparative et annoté: Provence et Languedoc. https://zenodo.org/records/15300719 (last accessed 31.10.2025).
Corbett, Greville G. 2006. Agreement. Cambridge: Cambridge Universitiy Press.
Cotterell, Ryan & Kirov, Christo & Hulden, Mans & Eisner, Jason. 2018. On the Diachronic Stability of Irregularity in Inflectional Morphology. arXiv:1804.08262. https://doi.org/10.48550/arXiv.1804.08262
Cucerzan, Silviu & Yarowsky, David. 2003. Minimally Supervised Induction of Grammatical Gender. In Proceedings of the 2003 Human Language Technology Conference of the North American Chapter of the Association for Computational Linguistics, 40–47. https://aclanthology.org/N03-1006/ DOI: https://doi.org/10.3115/1073445.1073451
dDOM = Stimm, Helmut & Stempel, Wolf-Dieter & Selig, Maria (eds.).1960–. Digitaler Zettelkasten des Dictionnaire de l’occitan medieval, München: Bayerische Akademie der Wissenschaften. https://dienste.badw.de:9999/dom/db (last accessed 31.10.2025).
Derrer, Felix. 1974. Lo codi : eine Summa codicis in provenzalischer Sprache aus dem XII. Jahrhundert; die provenzalische Fassung der Handschrift A; (Sorbonne 632); Vorarbeiten zu einer kritischen Textausgabe. Zürich: Juris.
DOMél = Stimm, Helmut & Stempel, Wolf-Dieter & Selig, Maria (eds.). 1960–. Dictionnaire de l’occitan medieval, Tübingen: Niemeyer. http://www.dom-en-ligne.de/ (last accessed 31.10.2025).
Donaldson, Bryan & Sibille, Jean. 2024. 8 Histoire interne de la langue. In Esher, Louise & Sibille, Jean (eds.). Manuel de linguistique occitane, 193–230. Berlin, Boston: De Gruyter. DOI: https://doi.org/10.1515/9783110733433-009
FEW = Wartburg, Walter Von. 1922–1987. Französisches Etymologisches Wörterbuch. Eine darstellung des galloromanischen Sprachschatzes, 25 vols., Leipzig/Bonn/Bâle: Teubner/Klopp/Zbinden. https://lecteur-few.atilf.fr/
Hare, Mary & Elman, Jeffrey L. 1995. Learning and morphological change. Cognition 56, 61–98. https://www.sciencedirect.com/science/article/pii/0010027794006555 DOI: https://doi.org/10.1016/0010-0277(94)00655-5
Hochreiter, Sepp & Schmidhuber, Jürgen. 1997. Long short-term memory. In Neural computation 9 (8), 1735–1780. https://www.bioinf.jku.at/publications/older/2604.pdf DOI: https://doi.org/10.1162/neco.1997.9.8.1735
Hockett, Charles Francis. 1962. A course in modern linguistics (4th edition). New York: Macmillan.
Jain, Sarthak & Wallace, Byron C. 2019. Attention is not Explanation. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), 3543–3556. https://doi.org/10.48550/arXiv.1902.10186
Jensen, Frede. 1973. Désaccord entre genre et flexion: Les substantifs masculins à desinence féminine en provençal. Revue des langues romanes 80: 393–404.
Jensen, Frede. 1976. The Old Provençal noun and adjective declension. Odense: Odense University Press.
Loporcaro, Michele. 2015. Vowel length from Latin to Romance. Oxford: Oxford University Press. DOI: https://doi.org/10.1093/acprof:oso/9780199656554.001.0001
Loporcaro, Michele. 2018. Gender from Latin to Romance history, geography, typology. Oxford: Oxford University Press. DOI: https://doi.org/10.1093/oso/9780199656547.003.0007
Marzo, Daniela & Wiedner, Marinus. 2025. Remarks on Grammatical Gender in Romance. In Linzmeier, Laura & Teixera Kalkhoff, Alexander M. & Wiesinger, Evelyn (eds.): Parla, e sia breve e arguto. Festschrift für Maria Selig / Studies in honor of Maria Selig (ScriptOralia 147). Tübingen: Narr, 201–207.
Panhuis, Dirk. 2015. Lateinische Grammatik. Berlin/Boston: De Gruyter. DOI: https://doi.org/10.1515/9783110405705
Pinkster, Harm. 2015. The Oxford Latin Syntax Volume I. Oxford: Oxford University Press. DOI: https://doi.org/10.1093/acprof:oso/9780199283613.001.0001
Polinsky, Maria & van Everbroeck, Ezra. 2003. Development of Gender Classifications: Modeling the Historical Change from Latin to French. Language 79(2): 356–390. https://www.jstor.org/stable/4489422?seq=1 DOI: https://doi.org/10.1353/lan.2003.0131
Sahai, Saumya & Sharma, Dravyansh. 2021. Predicting and Explaining French Grammatical Gender. In Proceedings of the Third Workshop on Computational Typology and Multilingual NLP, 90–96. https://aclanthology.org/2021.sigtyp-1.9/ DOI: https://doi.org/10.18653/v1/2021.sigtyp-1.9
Schöffel, Matthias & Wiedner, Marinus & Garcés Arias, Esteban & Ruppert, Paula & Heumann, Christian & Aßenmacher, Matthias. 2025. Modern Models, Medieval Texts: A POS Tagging Study of Old Occitan. In Hämäläinen, Mika & Öhman, Emily & Bizzoni, Yuri & Miyagawa, So & Alnajjar, Khalid (eds.), Proceedings of the 5th International Conference on Natural Language Processing for Digital Humanities, 334–349. https://aclanthology.org/2025.nlp4dh-1.30/ DOI: https://doi.org/10.18653/v1/2025.nlp4dh-1.30
Schöffel, Matthias & Garcés Arias, Esteban & Wiedner, Marinus & Ruppert, Paula & Li, Meimingwei & Heumann, Christian & Aßenmacher, Matthias. 2025. Unveiling Factors for Enhanced POS Tagging: A Study of Low-Resource Medieval Romance Languages. arXiv 2506.17715. https://doi.org/10.48550/arXiv.2506.17715
Serrano, Sofia & Smith, Noah A. 2019. Is Attention Interpretable?. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2931–2951. https://aclanthology.org/P19-1282/ DOI: https://doi.org/10.18653/v1/P19-1282
ThLL = Thesaurus linguae latinae. Berlin/Boston: De Gruyter. https://publikationen.badw.de/de/thesaurus/lemmata (last accessed 31.10.2025).
Vaswani, Ashish & Shazeer, Noam & Parmar, Niki & Uszkoreit, Jakob & Jones, Llion & Gomez, Aidan N. & Kaiser, Łukasz & Polosukhin, Illia. 2017. Attention Is All You Need. In Advances in Neural Information Processing Systems 30, 5998–6008. https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf
Veeman, Hartger & Basirat, Ali. 2020. An exploration of the encoding of grammatical gender in word embeddings. arXiv:2008.01946. https://doi.org/10.48550/arXiv.2008.01946
Wang, Ziwen. 2023. Pérdida del género neutro del latín al hispanorromance altomedieval. Una reconstrucción panrománica. Barcelona: Tesi. https://ddd.uab.cat/pub/tesis/2024/hdl_10803_691297/ziwa1de1.pdf
Wichers Schreur, Jesse & Allassonnière-Tang, Marc & Bellamy, Kate & Rochant, Neige. 2022. Predicting Grammatical Gender in Nakh Languages: Three Methods Compared. Linguistic Typology at the Crossroads 2 (2), 93–126. https://doi.org/10.6092/issn.2785-0943/14545
Wiedner, Marinus (2023): Old Occitan handwriting. (Modell-Nr. 52822, CER=3,51%), PyLaia-Modell for handwritten Occitan from the 13th and 14th century. [Computer software] https://readcoop.eu/model/oldoccitan-handwriting/ (last accessed 31.10.2025).
Wiedner, Marinus. Forthcoming. Doublons de genre en occitan médiéval : huit études de cas sur corpus. In Dufter, Andreas & Wissner, Inka (eds.): La variation en diachronie : regards sur la Galloromania (Beihefte zur Zeitschrift für romanische Philologie). Berlin/Boston: De Gruyter.
Wiedner, Marinus. Submitted. A Corpus Study on Grammatical Gender in Old Occitan. Freiburg im Breisgau: Inauguraldissertation an der Universität Freiburg.
Wiegreffe, Sarah & Pinter, Yuval. 2019. Attention is not not Explanation. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 11–20. https://aclanthology.org/D19-1002/ DOI: https://doi.org/10.18653/v1/D19-1002
Williams, Adina & Pimentel, Tiago & Blix, Hagen & McCarthy, Arya D. & Chodroff, Eleanor & Cotterell, Ryan. 2020. Predicting Declension Class from Form and Meaning. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 6682–6695. https://aclanthology.org/2020.acl-main.597/ DOI: https://doi.org/10.18653/v1/2020.acl-main.597
Details
License
Copyright (c) 2026 Marinus Wiedner, Matthias Schöffel
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.