Elamur Research⌕
☰
home / papers / 2026 / The evolution of linguistic patterns in scientific writing in Portuguese: an analysis with NILC-Metrix

The evolution of linguistic patterns in scientific writing in Portuguese: an analysis with NILC-Metrix

Original (Portuguese): Evolução de Padrões Linguísticos na Escrita Científica em Português: Uma Análise com NILC-Metrix

Paper · 2026 · Thiago Ruiz Lobo and Claudia Aparecida Martins

Summary

Large language models are increasingly used to draft and polish academic texts. This exploratory study asks whether that shift shows up in Portuguese scientific abstracts published by the Sociedade Brasileira de Computação (SBC). The authors scraped the SOL-LIB repository and filtered the records to 10,066 Portuguese abstracts from 2020 to 2025. They then computed 72 NILC-Metrix linguistic indicators for each abstract.

Using Welch's t-test, they compared each year from 2023 to 2025 against a pooled pre-LLM baseline (2020–2022). In 2023 most differences were not significant. In 2024 and 2025, highly significant differences dominated, and most of them pointed downward. Overall, abstracts became shorter and syntactically simpler, with weaker referential and semantic cohesion. At the same time, they became lexically denser, with more nouns and adjectives and fewer pronouns and function words.

The authors read this as a possible new pattern in Brazilian academic writing linked to generative AI. However, the study measures change over time and does not establish that AI caused it. The work offers a Portuguese-language baseline for future research on detecting AI-influenced text.

Key points

  • Data pipeline: 34,318 scraped SOL-LIB records were cleaned to 29,678, of which langdetect identified 17,490 as Portuguese (1987–2025). The 2020–2025 subset contains 10,066 abstracts.
  • Metrics: 72 NILC-Metrix metrics from the run_minimal script, covering 11 of the tool's 14 linguistic groups. The full set of about 200 metrics (run_all) could not be run for technical reasons. Psycholinguistic, textual simplicity and syntactic-pattern-density measures are absent.
  • Statistical method: Welch t-test with p < 0.05 treated as significant and p < 0.001 as highly significant. The sign of t indicates the direction of change.
  • Share of articles affected: No 2020–2022 article reaches any threshold. From 2023 onward:
    • over 95% of articles show at least 10 highly significant metric shifts;
    • about 92–95% show at least 15;
    • about 70–74% show at least 20;
    • 19–29% show at least 25;
    • 3–4% show at least 27.
  • Changes by group:
    • Syntactic complexity and referential cohesion fell across all their metrics.
    • Semantic cohesion dropped in 10 of 11 metrics in 2024 and 7 in 2025.
    • Lexical diversity rose in 2024–2025.
    • Brunet and Flesch scores declined, while Honore rose in 2025.
    • Connectives changed significantly but without a single direction.

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, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 works

Cite

Cite as Thiago Ruiz Lobo and Claudia Aparecida Martins. 2026. Evolução de Padrões Linguísticos na Escrita Científica em Português: Uma Análise com NILC-Metrix. Proceedings of the 17th International Conference on Computational Processing of Portuguese (PROPOR 2026) - Vol. 1.

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

@inproceedings{lobo2026evolucao,
  title = {{Evolução de Padrões Linguísticos na Escrita Científica em Português: Uma Análise com NILC-Metrix}},
  author = {Thiago Ruiz Lobo and Claudia Aparecida Martins},
  booktitle = {Proceedings of the 17th International Conference on Computational Processing of Portuguese (PROPOR 2026) - Vol. 1},
  year = {2026},
  url = {https://aclanthology.org/2026.propor-1.111/}
}

References

15 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Simone Astarita, Sandor Kruk, Jan Reerink, and Pablo Gómez. 2024. Delving into the utilisation of ChatGPT in scientific publications in astronomy. Proceedings of SPAICE2024: The First Joint European Space Agency / IAA Conference on AI in and for Space, pages 241–246. find
  2. Sergei Bagdasarov and Diego Alves. 2025. Like a human? a linguistic analysis of human-written and machine-generated scientific texts. In Proceedings of the First on Natural Language Processing and Language Models for Digital Humanities, pages 38–47, Varna, Bulgaria. INCOMA Ltd., Shoumen, Bulgaria. find
  3. Lorrie Cranor, Kim Hazelwood, Daniel Lopresti, and Amanda Stent. 2024. Conference submission and review policies to foster responsible computing research. arXiv preprint arXiv:2408.09678. — [arXiv:2408.09678](https://arxiv.org/abs/2408.09678) find
  4. Bernard L. Welch. 1947. The generalization of “student’s” problem when several different population variances are involved. Biometrika, 34(1–2):28–35. find
  5. Sergio E. Zanotto and Segun Aroyehun. 2024. Human variability vs. machine consistency: A linguistic analysis of texts generated by humans and large language models. arXiv preprint arXiv:2412.03025. — [arXiv:2412.03025](https://arxiv.org/abs/2412.03025) find
  6. Lia Cornelia Culda, Raluca Andreea Nerişanu, Marian Pompiliu Cristescu, Dumitru Alexandru Mara, Adela Bâra, and Simona-Vasilica Oprea. 2025. Comparative linguistic analysis framework of human-written vs. machine-generated text. Connection Science, 37(1):2507183. find
  7. Mingmeng Geng and Roberto Trotta. 2024. Is ChatGPT transforming academics’ writing style? ICML 2024 Next Generation of AI Safety Workshop. find
  8. Georgios P. Georgiou. 2025. Differentiating between human-written and AI-generated texts using automatically extracted linguistic features. Information, 16(11):979. find
  9. Dmitry Kobak, Rita González-Márquez, Emőke Ágnes Horvát, and Jan Lause. 2025. Delving into LLM-assisted writing in biomedical publications through excess vocabulary. Science Advances, 11(27):eadt3813. find
  10. Nataliya Kosmyna, Eugene Hauptmann, Ye Tong Yuan, Jessica Situ, Xian-Hao Liao, Ashly Vivian Beres-nitzky, Iris Braunstein, and Pattie Maes. 2025. Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. arXiv preprint arXiv:2506.08872. — [arXiv:2506.08872](https://arxiv.org/abs/2506.08872) find
  11. S. E. Leal, M. S. Duran, C. E. Scarton, and 1 others. 2024. NILC-Metrix: assessing the complexity of written and spoken language in brazilian portuguese. Language Resources and Evaluation, 58:73–110. find
  12. Weixin Liang, Yaohui Zhang, Zhengxuan Wu, Haley Lepp, Wenlong Ji, Xuandong Zhao, Hancheng Cao, Sheng Liu, Siyu He, Zhi Huang, Diyi Yang, Christopher Potts, Christopher D. Manning, and James Y. Zou. 2024. Mapping the increasing use of LLMs in scientific papers. arXiv preprint arXiv:2404.01268.
  13. Dingkang Lin, Naixuan Zhao, Dan Tian, and Jiang Li. 2025. ChatGPT as linguistic equalizer? quantifying LLM-driven lexical shifts in academic writing. arXiv preprint arXiv:2504.12317. — [arXiv:2504.12317](https://arxiv.org/abs/2504.12317) find
  14. A. Muñoz-Ortiz, C. Gómez-Rodríguez, and D. Vilares. 2024. Contrasting linguistic patterns in human and LLM-generated news text. Artificial Intelligence Review, 57:265. find
  15. Flávia A. Rodrigues, Niclas F. Sturm, and Flávio L. Pinheiro. 2026. A linguistic comparison between human- and AI-generated content. iScience, 29(3). find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/lobo-2026-evolucao-de-padroes-linguisticos-na/