Elamur Research⌕
☰
home / papers / 2026 / Parsing Nheengatu: Performance Gains for a Brazilian Indigenous Universal Dependencies Treebank

Parsing Nheengatu: Performance Gains for a Brazilian Indigenous Universal Dependencies Treebank

Summary

The authors test whether expanding UD_Nheengatu-CompLin improves the quality of automatic syntactic parsing. It is the only UD treebank of Nheengatu, an endangered Indigenous language of Brazil. The corpus was expanded with 19th-century records of the Lower Amazon language collected by Charles Hartt. These data introduced archaic morphosyntactic phenomena: old imperative forms, SOV word order and a dative pronoun. The researchers compared two versions of the corpus: one without the Hartt material and one with it. For this they used UDPipe 1.4.0 and 10-fold cross-validation in two settings: with gold tokenization and gold tags, and with fully automatic analysis of raw text. Statistical significance was tested with the Mann–Whitney test. The authors hoped for a LAS gain of at least 10%, but did not obtain it. However, with the expanded corpus accuracy increased and results across folds became more stable. In the gold-input setting the difference is statistically significant. The work shows that for low-resource languages, incremental data growth, consistent annotation and a reproducible evaluation methodology matter.

Key points

  • In the UD 2.17 release the treebank contains 2,742 trees and 26,033 words. It has a 3.5-star rating, whereas none of the other 24 treebanks of Indigenous languages of the Americas has more than 2.
  • Experiment 1 used nohartt.conllu (2,091 sentences). Experiment 2 used all.conllu (2,834 sentences), of which 743 are taken from Hartt (1938). The data are published in the CompLin/nheengatu repository on GitHub.
  • Aggregate evaluation over all test splits (26,785 tokens): LAS 84.28%, UAS 88.11%, label accuracy 90.82%.
  • Head-only errors (1,750) are more frequent than label-only errors (1,026). Another 1,434 cases have both errors. The parser most often confuses nsubj with obj, and also root with parataxis. It often replaces the labels advcl, ccomp and xcomp with parataxis or root.
  • In the automatic setting mean accuracy barely changed, but the variance across folds decreased. Historical texts have more orthographic variants, so morphosyntactic tagging on them is harder.

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

Topics: Universal Dependencies · Treebanks

Cite

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{alexandre2026parsing,
  title = {{Parsing Nheengatu: Performance Gains for a Brazilian Indigenous Universal Dependencies Treebank}},
  author = {Alexandre, Dominick Maia and Alencar, Leonel Figueiredo de},
  booktitle = {Proceedings of the 17th International Conference on Computational Processing of Portuguese (PROPOR 2026) - Vol. 2},
  year = {2026},
  url = {https://aclanthology.org/2026.propor-2.29/}
}

References

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

  1. Alexandra Y. Aikhenvald and R. M. W. Dixon. 2001. Introduction. In Alexandra Y. Aikhenvald and R. M. W. Dixon, editors, Areal diffusion and genetic inheritance: Problems in comparative linguistics, pages 1–26. Oxford University Press, Oxford. find
  2. Leonel Figueiredo de Alencar. 2024. UD_Nheengatu-CompLin: o corpus sintaticamente anotado do nheengatu da coleção Universal Dependencies. In Anais Eletrônicos do XVI Encontro de Linguística de Corpus e da XII Escola Brasileira de Linguística Computacional, volume 1, pages 105–109, Brasília. As-socia… find
  3. Leonel Figueiredo de Alencar. 2024b. Aspectos da construção de um corpus sintaticamente anotado do nheengatu no modelo Dependências Universais. Texto Livre, 17:e52653. find
  4. Dominick Maia Alexandre and Leonel Figueiredo de Alencar. 2025. Universal Dependencies for 19th-Century Nheengatu from the Lower Amazon Region. In Anais do XVI Simpósio Brasileiro de Tecnologia da Informação e da Linguagem Humana, pages 588– 598, Porto Alegre, RS, Brasil. SBC. find
  5. Leonel Figueiredo de Alencar. 2025. Enhancing a Nheengatu Morphosyntactic Analyzer for Word Formation and Non-standard Language. In Anais do XVI Simpósio Brasileiro de Tecnologia da Informação e da Linguagem Humana, pages 13–28, Porto Alegre, RS, Brasil. SBC. find
  6. Marcel Twardowsky Avila. 2021. Proposta de dicionário nheengatu-português. Ph.D. thesis, Faculdade de Filosofia, Letras e Ciências Humanas da Universidade de São Paulo. find
  7. Magali Duran, Lucelene Lopes, Maria das Graças Nunes, and Thiago Pardo. 2023. The Dawn of the Porttinari Multigenre Treebank: Introducing its Journalistic Portion. In Anais do XIV Simpósio Brasileiro de Tecnologia da Informação e da Linguagem Humana, pages 115–124, Porto Alegre, RS, Brasil. SBC. find
  8. S. Bird. 2020. Decolonising speech and language technology. In COLING 2020 - 28th International Conference on Computational Linguistics, Proceedings of the Conference, COLING 2020 - 28th International Conference on Computational Linguistics, Proceedings of the Conference, pages 3504–3519. Associatio… find
  9. Frederic Blum. 2022. Evaluating zero-shot transfers and multilingual models for dependency parsing and POS tagging within the low-resource language family tupían. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop, pages 1–9, Dublin,… find
  10. Luiz Carlos Borges. 1996. O nheengatú: uma língua amazônica. Papia, 4(2):44–55. find
  11. Kenneth Church and Mark Liberman. 2021. The future of computational linguistics: On beyond alchemy. Frontiers in Artificial Intelligence, 4:1–18. find
  12. Aline da Cruz. 2011. Fonologia e gramática do nheengatú: A língua falada pelos povos Baré, Warekena e Baniwa. LOT, Utrecht. find
  13. Diego Pedro Gonçalves da Silva and Thiago Alexandre Salgueiro Pardo. 2024. Grammar induction for Brazilian indigenous languages. In Proceedings of the 16th International Conference on Computational Processing of Portuguese - Vol. 2, pages 64–72, Santiago de Compostela, Galicia/Spain. Association for… find
  14. Leonel Figueiredo de Alencar. 2023. Yauti: A tool for morphosyntactic analysis of Nheengatu within the Universal Dependencies framework. In Anais do XIV Simpósio Brasileiro de Tecnologia da Informação e da Linguagem Humana, pages 135–145, Porto Alegre, RS, Brasil. SBC. find our card
  15. Leonel Figueiredo de Alencar. 2024a. A Universal Dependencies Treebank for Nheengatu. In Proceedings of the 16th International Conference on Computational Processing of Portuguese, volume 2, pages 37–54, Santiago de Compostela, Galicia, Spain. Association for Computational Linguistics. find our card
  16. David M. Eberhard, Gary F. Simons, and Charles D. Fennig, editors. 2025. Ethnologue: Languages of the World, 28 edition. SIL International, Dallas. find
  17. Charles Frederick Hartt. 1872. Notes on the Lingoa Geral or Modern Tupi of the Amazonas. Transactions of the American Philological Association, 3:58– 76. find
  18. Charles Frederick Hartt. 1938. Notas sobre a língua geral, ou tupí moderno do Amazonas. Anais da Biblioteca Nacional do Rio de Janeiro, LI:305–390. [1929]. find
  19. Pratik Joshi, Sebastin Santy, Amar Budhiraja, Kalika Bali, and Monojit Choudhury. 2020. The State and Fate of Linguistic Diversity and Inclusion in the NLP World. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6282–6293, Online. Association for Comp… find our card
  20. Lucelene Lopes and Thiago Pardo. 2024. Towards Portparser - a highly accurate parsing system for Brazilian Portuguese following the Universal Dependencies framework. In Proceedings of the 16th International Conference on Computational Processing of Portuguese - Vol. 1, pages 401–410, Santiago de Com… find
  21. Lucelene Lopes, Thiago Pardo, and Magali Duran. 2024. Syntactic parsing: where are we going? In Anais do XV Simpósio Brasileiro de Tecnologia da Informação e da Linguagem Humana (STIL 2024), pages 67–74, Porto Alegre, RS, Brasil. SBC. find
  22. Manuel Mager, Ximena Gutierrez-Vasques, Gerardo Sierra, and Ivan Meza-Ruiz. 2018. Challenges of language technologies for the indigenous languages of the Americas. In Proceedings of the 27th International Conference on Computational Linguistics, pages 55–69, Santa Fe, New Mexico, USA. Association fo… find
  23. Henry B. Mann and Donald R. Whitney. 1947. On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other. The Annals of Mathematical Statistics, 18(1):50–60. find
  24. Lorena Martín Rodríguez and 1 others. 2022. Tupían language ressources: Data, tools, analyses. In Proceedings of the 1st Annual Meeting of the ELRA/ISCA Special Interest Group on Under-Resourced Languages, pages 48–58, Marseille, France. European Language Resources Association. find
  25. Denny Moore. 2014. Historical development of Nheengatu (Língua Geral Amazônica). In Salikoko S. Mufwene, editor, Iberian Imperialism and Language Evolution in Latin America, pages 108–142. University of Chicago Press, Chicago. find
  26. Denny Moore, Sidney Facundes, and Nádia Pires. 1994. Nheengatu (Língua Geral Amazônica), its history, and the effects of language contact. In Proceedings of the Meeting of the Society for the Study of the Indigenous languages of the Americas, July 2-4, 1993 and the Hokan-Penutian Workshop, July 3, 1… find
  27. Eduardo de Almeida Navarro. 2012. O último refú-gio da língua geral no Brasil. Estudos Avançados, 26(76):245–254. find
  28. Eduardo de Almeida Navarro, Marcel Twardowsky Ávila, and Rodrigo Godinho Trevisan. 2017. O Nheengatu, entre a vida e a morte: A tradução literária como possível instrumento de sua revitalização lexical. Revista Letras Raras, 6(2):9–29. find
  29. Joakim Nivre, Daniel Zeman, Filip Ginter, and Francis Tyers. 2017. Universal Dependencies. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Tutorial Abstracts, Valencia, Spain. Association for Computational Linguistics. find
  30. Milan Straka, Jan Hajič, and Jana Straková. 2016. UD-Pipe: Trainable pipeline for processing CoNLL-U files performing tokenization, morphological analysis, POS tagging and parsing. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16), pages 4290– 4297,… find
  31. Guillaume Thomas. 2019. Universal Dependencies for Mbyá Guaraní. In Proceedings of the Third Workshop on Universal Dependencies (UDW, SyntaxFest 2019), pages 70–77, Paris, France. Association for Computational Linguistics. find
  32. Francis M. Tyers and Robert Henderson. 2021. A corpus of K’iche’ annotated for morphosyntactic structure. In Proceedings of the First Workshop on NLP for Indigenous Languages of the Americas (AmericasNLP). find
  33. Alonso Vasquez and 1 others. 2018. Toward Universal Dependencies for Shipibo-konibo. In Proceedings of the Second Workshop on Universal Dependencies (UDW 2018), pages 151–161, Brussels, Belgium. Association for Computational Linguistics. find
  34. Pauli Virtanen, Ralf Gommers, Travis E. Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Ev-geni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, Stéfan J. van der Walt, Matthew Brett, Joshua Wilson, K. Jarrod Millman, Nikolay Mayorov, Andrew R. J. Nelson, Eric Jones, Robert Kern… find our card
  35. Robert Pugh, Marivel Huerta Mendez, Mitsuya Sasaki, and Francis Tyers. 2022. Universal Dependencies for Western Sierra Puebla Nahuatl. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 5011–5020, Marseille, France. European Language Resources Association. find
  36. Aryon Dall’Igna Rodrigues. 1986. Línguas Brasileiras: Para o conhecimento das línguas indígenas. Loyola, São Paulo. Vários quadros numerados e outros sem numeração. find
  37. Elvis Souza, Aline Silveira, Tatiana Cavalcanti, Maria Castro, and Claudia Freitas. 2021. PetroGold – Corpus padrão ouro para o dominio do petroleo. In Proceedings of the 13th Brazilian Symposium in Information and Human Language Technology, pages 29–38, Porto Alegre, Brazil. Association for Computa… find
  38. Ermanno Stradelli. 2014. Vocabulário português-nheengatu, nheengatu-português. Ateliê Editorial, Cotia, SP. Original work published in 1929. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/alexandre-2026-parsing-nheengatu-performance-gains-brazilian/