Elamur Research⌕
☰
home / papers / 2026 / The Visibility of Depression in Social Media: Mapping Symptoms to Linguistic Features

The Visibility of Depression in Social Media: Mapping Symptoms to Linguistic Features

Paper · 2026 · Ștefana-Arina Tăbușcă, Ana Sabina Uban and Liviu Dinu

Summary

Text-based depression detection usually treats depression as one target. This leaves open which individual symptoms actually appear in what people write online. The authors address this with eRisk 2021, which pairs BDI-II questionnaire answers with Reddit posting histories. After restricting to English content, the sample is 169 users with 72,430 posts. They first model the 21 BDI-II items as a Gaussian Graphical Model (EBICglasso, polychoric correlations). A sanity check with bge-large-en-v1.5 embeddings confirms that the posts carry a depression signal. Each item is then correlated with 15 hand-picked LIWC-22 categories, giving a 21 × 15 symptom-language bridge matrix with Benjamini-Hochberg FDR correction. Only 37 of 315 pairs remain significant. Cognitive and affective symptoms such as worthlessness, sadness and suicidality are tied to terms about mental health, expressions of anxiety and self-references in the first person singular. Sleep, appetite, irritability and libido changes show no significant links. This suggests that common LIWC-style monitoring may overlook bodily dimensions of depression. The authors acknowledge that the feature selection and the small sample limit this conclusion.

Key points

  • Data: eRisk 2021, 169 Reddit users with BDI-II scores, 72,430 English posts (median 282 per user). An embedding-based severity proxy correlated with the BDI-II total at ρ = 0.463.
  • Clinical network: 100 non-zero edges, density 0.476, expected-influence stability CS = 0.438. Worthlessness (Q14) is the most central node (1.321), while libido loss (0.450) and appetite changes (0.590) are the least central. Walktrap found 5 symptom communities.
  • Bridge matrix: 37 significant associations involve 17 of 21 symptoms and 7 of 15 LIWC features. Mental health vocabulary accounts for 13 associations, anxiety words for 11 and first-person pronouns for 8. The only negative link is conflict words with past failure (ρ = −0.219).
  • At the community level, 21 of 75 associations are significant. Cognitive-affective communities surface through self-referential language, while somatic-vegetative and motivational communities surface through illness vocabulary.
  • The authors list several limitations: self-selected Reddit users, self-reported scores, word-level LIWC that ignores context, and wide confidence intervals with N = 169. They propose contextual embeddings and cross-platform replication as next steps.

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

 Institutions

🎓University of BucharestBucharest · Romania21 works · 516 citations

 Fields, topics and concepts

📚 Shelves 3The field of science this work belongs to (OpenAlex fields).Psychology · 1,552 worksExperimental and Cognitive Psychology · 870 worksSocial Psychology · 306 works

Cite

Cite as Ștefana-Arina Tăbușcă, Ana Sabina Uban and Liviu Dinu. 2026. The Visibility of Depression in Social Media: Mapping Symptoms to Linguistic Features. Proceedings of the 11th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.clpsych-1.28

@inproceedings{tabusca2026visibility,
  title = {{The Visibility of Depression in Social Media: Mapping Symptoms to Linguistic Features}},
  author = {Ștefana-Arina Tăbușcă, Ana Sabina Uban and Liviu Dinu},
  booktitle = {Proceedings of the 11th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.clpsych-1.28},
  url = {https://doi.org/10.18653/v1/2026.clpsych-1.28}
}

References

21 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Aaron T Beck, Robert A Steer, Gregory K Brown, and 1 others. 1996. Beck depression inventory. find
  2. Denny Borsboom. 2017. A network theory of mental disorders. World Psychiatry, 16(1):5–13. find
  3. Ryan L Boyd, Ashwini Ashokkumar, Sarah Seraj, and James W Pennebaker. 2022. The development and psychometric properties of liwc-22. Austin, TX: University of Texas at Austin, 10. find
  4. Ana-Maria Bucur, Ioana R. Podină, and Liviu P. Dinu. 2021. A Psychologically Informed Part-of-Speech Analysis of Depression in Social Media. In Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021), pages 199–207. find
  5. Sacha Epskamp, Denny Borsboom, and Eiko I. Fried. 2018. Estimating psychological networks and their accuracy: A tutorial paper. Behavior Research Methods, 50(1):195–212. find
  6. Rina Foygel and Mathias Drton. 2010. Extended bayesian information criteria for gaussian graphical models. Preprint, arXiv:1011.6640. — [arXiv:1011.6640](https://arxiv.org/abs/1011.6640) find
  7. Robert M. A. Hirschfeld. 2001. The comorbidity of major depression and anxiety disorders: Recognition and management in primary care. Primary Care Companion for CNS Disorders, 3(6):24412. find
  8. Javier Parapar, Patricia Martín-Rodilla, David E Losada, and Fabio Crestani. 2021. Overview of erisk at clef 2021: Early risk prediction on the internet (extended overview). CLEF (Working Notes), 1:864–887. find
  9. Donald Robinaugh, Alexander Millner, and Richard Mc-Nally. 2016. Identifying highly influential nodes in the complicated grief network. Journal of Abnormal Psychology, 125:747–757. find
  10. Stephanie Rude, Eva-Maria Gortner, and James W. Pennebaker. 2004. Language use of depressed and depression-vulnerable college students. Cognition & Emotion, 18(8):1121–1133. find
  11. Marten Scheffer, Claudi Bockting, Denny Borsboom, Roshan Cools, Clara Delecroix, Jessica Hartmann, Kenneth Kendler, Ingrid van de Leemput, Han Maas, Egbert Nes, Mark Mattson, Pat McGorry, and Barnaby Nelson. 2024. A dynamical systems view of psychiatric disorders—theory: A review. JAMA Psychiatry, 8… find
  12. Jelena Spasojević and Lauren Alloy. 2001. Rumination as a common mechanism relating depressive risk factors to depression. Emotion (Washington, D.C.), 1:25–37. find
  13. Katrine Bønneland Tølbøll. 2019. Linguistic features in depression: a meta-analysis. find
  14. Laura Bringmann, Lotte Lemmens, Marcus Huibers, Denny Borsboom, and Francis Tuerlinckx. 2014. Revealing the dynamic network structure of the beck depression inventory-ii. Psychological Medicine, pages 1–11. find
  15. Ana-Sabina Uban, Berta Chulvi, and Paolo Rosso. 2021. An Emotion and Cognitive Based Analysis of Mental Health Disorders from Social Media Data. Future Generation Computer Systems, 124:480–494. find
  16. Ana-Maria Bucur, Adrian Moldovan, Keerthi Parvatikar, Marcos Zampieri, Ashiqur R. KhudaBukhsh, and Liviu P. Dinu. 2025. On the State of NLP Approaches to Modeling Depression in Social Media: A Post-COVID-19 Outlook. IEEE Journal of Biomedical and Health Informatics, 29(6):4439–4451. find
  17. Ana-Sabina Uban, Berta Chulvi, and Paolo Rosso. 2022. Multi-Aspect Transfer Learning for Detecting Low Resource Mental Disorders on Social Media. In Proceedings of the Thirteenth Language Resources and Evaluation Conference (LREC 2022), pages 3202– 3219. find
  18. Claudia van Borkulo, Lynn Boschloo, Denny Borsboom, B.W. Penninx, Lourens Waldorp, and Robert Schoevers. 2015. Association of symptom network structure with the course of depression. JAMA Psychiatry, 72. find
  19. Hongli Zhan, Desmond Ong, and Junyi Jessy Li. 2023. Evaluating subjective cognitive appraisals of emotions from large language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 14418–14446. find
  20. Tianlin Zhang, Kailai Yang, Shaoxiong Ji, and Sophia Ananiadou. 2023. Emotion fusion for mental illness detection from social media: A survey. Information Fusion, 92:231–246. find
  21. Hamad Zogan, Imran Razzak, Xianzhi Wang, Shoaib Jameel, and Guandong Xu. 2022. Explainable depression detection with multi-aspect features using a hybrid deep learning model on social media. World Wide Web, 25(1):281–304. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tabusca-2026-visibility-depression-social-media-mapping/