Elamur Research⌕
☰
home / papers / 2026 / The visible and the latent linguistic clues of mental health in Brazilian Portuguese textual posts

The visible and the latent linguistic clues of mental health in Brazilian Portuguese textual posts

Paper · 2026 · Rodrigo Wilkens, Helena Caseli, Vania Neris et al.

Summary

Depression is common in Brazil, yet many people never receive care. Text-based screening could help, but signs of depression are expressed differently across languages and cultures. This study looks at how depressive symptoms appear in Brazilian Portuguese narrative posts. It uses the Amive corpus, which contains Facebook posts from pages where Brazilian college students share content. Mental health trainees annotated each post for possible depressive profile (PDP) and marked text spans for 18 symptoms. The authors refined the span boundaries and then compared PDP and non-PDP texts, and the symptom spans, on three kinds of features:

  • Lexical: tokens, types, TTR, syllables per word and a Portuguese-adapted Flesch score.
  • Syntactic: features including person marking and null subjects.
  • Psycholinguistic: lexicon-based measures.

They also fitted a Bayesian hierarchical model to link symptoms to PDP status. PDP texts differ measurably from non-PDP texts. PDP seems to intensify negative, rigid and self-focused language within symptoms, and suicidal ideation is the strongest predictor of PDP. These findings point to features that could inform more tailored Portuguese mental-health NLP tools.

Key points

  • The revised Amive corpus has 604 posts: 336 PDP and 268 non-PDP. They were selected by depression-related keywords and published between 2012 and 2021. Four students of psychology, psychiatry or occupational therapy did the annotation.
  • Within the Suicide/Self-extermination category, PDP posts use about twice as much death-related vocabulary (0.0223 vs 0.0100). In the Feeling of guilt category, PDP posts show more negative emotion (0.0477 vs 0.0346). PDP posts also tend to use more negation and absolutist words across most symptoms.
  • In the Bayesian model, suicidal ideation has an odds ratio for PDP of about 4.55 (β = 1.516). Despair has an odds ratio of about 2.30 and sadness about 2.14. Low self-esteem and functional impairment each come close to doubling the odds.
  • Irritation/Aggressiveness shows a negative association with PDP (OR ≈ 0.60) once the other symptoms are controlled for.
  • The authors note that the small corpus and the Facebook keyword-based sampling are limitations. Planned future work covers figurative language and an English replication within the UK–Brazil AIM-Health project.

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 Rodrigo Wilkens, Helena Caseli, Vania Neris and Aline Villavicencio. 2026. The visible and the latent linguistic clues of mental health in Brazilian Portuguese textual posts. Proceedings of the 17th International Conference on Computational Processing of Portuguese (PROPOR 2026) - Vol. 2.

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

@inproceedings{wilkens2026visible,
  title = {{The visible and the latent linguistic clues of mental health in Brazilian Portuguese textual posts}},
  author = {Rodrigo Wilkens, Helena Caseli, Vania Neris and Aline Villavicencio},
  booktitle = {Proceedings of the 17th International Conference on Computational Processing of Portuguese (PROPOR 2026) - Vol. 2},
  year = {2026},
  url = {https://aclanthology.org/2026.propor-2.21/}
}

References

12 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. David M. Blei, Andrew Y. Ng, and Michael I. Jordan. 2003. Latent dirichlet allocation. J. Mach. Learn. Res., 3(null):993–1022.
  2. Margaret M Bradley and Peter J Lang. 1999. Affective norms for english words (anew): Instruction manual and affective ratings. Technical report, Technical report C-1, The center for research in psychophysiology, University of Florida. find
  3. Vinicius Casani, Alinne C. Correa Souza, Rafael G. Mantovani, and Francisco Carlos M. Souza. 2021. DP-symptom-identifier: uma estratégia para classificar sintomas de depressão utilizando um conjunto de dados textuais na língua portuguesa. In Annals of the XIII Brazilian Symposium of Information Tecn… find
  4. Munmun De Choudhury, Michael Gamon, Scott Counts, and Eric Horvitz. 2013. Predicting depression via social media. In Proceedings of the international AAAI conference on web and social media, volume 7, pages 128–137. find
  5. Antonia Errazuriza, Dalia Avello-Vegab, Juan P. Ramirez-Mahalufa, Rafael Torresa, Nicolas A. Cross-leya, Eduardo A. Undurragac, and Peter B. Jones. 2023. Prevalence of depressive disorder in the adult population of latin america: a systematic review and meta-analysis. The Lancet Regional Health – Am… find
  6. Pedro P. Balage Filho, Thiago Alexandre Salgueiro Pardo, and Sandra M. Aluísio. 2013. An evaluation of the Brazilian Portuguese LIWC dictionary for sentiment analysis. In Proceedings of the 9th Brazilian Symposium in Information and Human Language Technology. find
  7. Global Burden of Disease Collaborative Network. 2025. Global Burden of Disease Study 2023 (GBD 2023) Covariates 1980-2023. Institute for Health Metrics and Evaluation (IHME), Seattle, United States of America. find
  8. Maarten Grootendorst. 2022. Bertopic: Neural topic modeling with a class-based tf-idf procedure. arXiv preprint arXiv:2203.05794.
  9. Shaoxiong Ji, Celina Ping Yu, Sai-fu Fung, Shirui Pan, Guodong Long, and Gao Cong. 2018. Supervised learning for suicidal ideation detection in online user content. find
  10. Paulo Mann, Aline Paes, and Elton H. Matsushima. 2020. See and read: Detecting depression symptoms in higher education students using multimodal social media data. Proceedings of the International AAAI Conference on Web and Social Media, 14:440–451. find
  11. Paulo Mann, Matheus Yasuo Ribeiro Utino, Elton Hiroshi Matsushima, and Aline Paes. 2025. The topics of depression on social networking sites. Journal of the Brazilian Computer Society, 31(1):771–806. find
  12. Teresa BF Martins, Claudete M Ghiraldelo, Maria das Graças Volpe Nunes, and Osvaldo Novais de Oliveira Junior. 1996. Readability formulas applied to textbooks in brazilian portuguese. – tk↑, ty↑ tk↑, ttr↓, ty↑ ttr↓ Symptom-Level Pairwise Linguistic Contrasts – tk↓, ty↓ – tk↓, ty↓ – tk↑, ty↑ tk↓, ttr… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wilkens-2026-visible-latent-linguistic-clues-mental/