Elamur Research⌕
☰
home / papers / 2026 / TONY: an open-source TOolkit for Nlp in psYchology

TONY: an open-source TOolkit for Nlp in psYchology

Paper · 2026 · Federico Ravenda, Sofia Irene Ravenda, Volodymyr Karpenko et al.

Summary

Clinical psychology increasingly wants language-based tools that scale, but available NLP resources tend to be scattered across single studies, tied to one condition, or closed-source, as with LIWC. TONY is an open-source Python package that combines two approaches. The first is an interpretable linguistic feature extractor that covers nine families of markers, including lexical, syntactic (via spaCy), morphosyntactic, stylistic, affective and cohesion features. It writes one row of features per document. The second is a set of lightweight transformer modules adapted to mental health. These estimate perplexity, predict HiTOP traits, interpret Sparse Autoencoder features, complete the BDI-II questionnaire, model topics, profile emotions, and flag interpersonal risk factors (IRF) and cognitive biases with supporting text spans.

The authors evaluated TONY on several Reddit-derived mental health datasets. Its features rivalled or beat LIWC and clearly beat Empath, and its fine-tuned and retrieval-based components outperformed earlier systems. The toolkit matters because it gives interdisciplinary researchers a single free and reproducible pipeline that avoids both expensive commercial software and very large models.

Key points

  • Feature comparison against LIWC and Empath:
    • On DepreSym, TONY's features beat LIWC on every metric except precision.
    • On the IRF thwarted-belongingness (TBE) task, they consistently beat both baselines.
    • On Depression Severity, they reached the highest accuracy.
  • IRF classifier: A 4-bit quantized Qwen-8B fine-tuned on the IRF training split outperformed Mental-BERT as well as zero-shot gpt-5.4-mini and Qwen-2.5-72B-instruct. The fine-tuned models are on Hugging Face under FritzStack/models with the "IRF" prefix.
  • BDIScorer: This class takes a user's Reddit post history and uses an adaptive RAG approach to predict a 21-item BDI-II score vector. On eRisk, every configuration outperformed DepressMind and the top eRisk 2021 participants.
  • Limitations: The models are trained mostly on English Reddit text, and the authors state the tool is for research only, not clinical diagnosis.
  • Funding: The work was partly funded by SNSF grant 200,557.

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 Milano-BicoccaMilan · Italy15 works · 4,946 citations🎓Università della Svizzera italianaLugano · Switzerland6 works · 647 citations🎓University of PaviaPavia · Italy6 works · 69 citations🎓University of InsubriaVarese · Italy3 works · 1 citation

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksSocial Psychology · 306 works

Cite

Cite as Federico Ravenda, Sofia Irene Ravenda, Volodymyr Karpenko, Daniele Montagnani, Andrea Raballo and Antonietta Mira. 2026. TONY: an open-source TOolkit for Nlp in psYchology. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-demo.65

@inproceedings{ravenda2026tony,
  title = {{TONY: an open-source TOolkit for Nlp in psYchology}},
  author = {Federico Ravenda, Sofia Irene Ravenda, Volodymyr Karpenko, Daniele Montagnani, Andrea Raballo and Antonietta Mira},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-demo.65},
  url = {https://doi.org/10.18653/v1/2026.acl-demo.65}
}

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. Aakash Kumar Agarwal, Saprativa Bhattacharjee, Mauli Rastogi, Jemima S. Jacob, Biplab Banerjee, Rashmi Gupta, and Pushpak Bhattacharyya. 2025. ReDepress: A cognitive framework for detecting depression relapse from social media. In Proceedings of the 2025 Conference on Empirical Methods in Natural La… find
  2. Stefano Baccianella, Andrea Esuli, Fabrizio Sebastiani, and 1 others. 2010. Sentiwordnet 3.0: an enhanced lexical resource for sentiment analysis and opinion mining. In Lrec, volume 10, pages 2200–2204. Valletta.
  3. Simar Bajaj. 2024. How students and grandparents could solve the global mental-health crisis. Nature, 635(8039):540–542. find
  4. Margaret M. Bradley and Peter J. Lang. 1999. Affective norms for english words (anew): Instruction manual and affective ratings. find
  5. Suchandra Chakraborty, Sudeshna Jana, Manjira Sinha, and Tirthankar Dasgupta. 2025. Self-state evidence extraction and well-being prediction from social media timelines. In Proceedings of the 10th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2025), pages 279–286. find
  6. Alessandro De Grandi, Federico Ravenda, Andrea Raballo, and Fabio Crestani. 2025. The emotional spectrum of llms: Leveraging empathy and emotion-based markers for mental health support. In Proceedings of the 10th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2025), pages 26–… find
  7. Oliver Delgaram-Nejad, Dawn Archer, Gerasimos Chatzidamianos, Louise Robinson, and Alex Bartha. 2023. The dais-c: A small, specialised, spoken, schizophrenia corpus. Applied Corpus Linguistics, 3(3):100069. find
  8. Antonio Di Noia, Federico Ravenda, and Antonietta Mira. 2026. A general framework for adaptive non-parametric dimensionality reduction. Scientific Reports. find
  9. Ethan Fast, Binbin Chen, and Michael S Bernstein. 2016. Empath: Understanding topic signals in large-scale text. In Proceedings of the 2016 CHI conference on human factors in computing systems, pages 4647– 4657. find
  10. Roque Fernández-Iglesias, Marcos Fernández-Pichel, Mario Ezra Aragón, and David E Losada. 2024. Depressmind: A depression surveillance system for social media analysis. In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstratio… find
  11. Rudolph Flesch. 1948. A new readability yardstick. Journal of applied psychology, 32(3):221. find
  12. Muskan Garg, Amirmohammad Shahbandegan, Amrit Chadha, and Vijay Mago. 2023. An annotated dataset for explainable interpersonal risk factors of mental disturbance in social media posts. In Findings of the Association for Computational Linguistics: ACL 2023, pages 11960–11969, Toronto, Canada. Associa… find
  13. Matej Gjurković, Vanja M Karan, Iva Vukojević, Mihaela Bošnjak, and Jan Šnajder. 2021. Pandora talks: Personality and demographics on reddit. In Proceedings of the ninth international workshop on natural language processing for social media, pages 138– 152. find
  14. Maarten Grootendorst. 2022. Bertopic: Neural topic modeling with a class-based tf-idf procedure. arXiv preprint arXiv:2203.05794.
  15. Zhijun Guo, Alvina Lai, Johan H Thygesen, Joseph Farrington, Thomas Keen, and Kezhi Li. 2024. Large language models for mental health applications: systematic review. JMIR mental health, 11(1):e57400. find
  16. Frank E Harrell Jr. 2015. Ordinal logistic regression. In Regression modeling strategies: with applications to linear models, logistic and ordinal regression, and survival analysis, pages 311–325. Springer. find
  17. Jiaman He, Marta Micheli, Damiano Spina, Dana McKay, Johanne R Trippas, and Noriko Kando. 2026. Characterizing personality from eye-tracking: The role of gaze and its absence in interactive search environments. arXiv preprint arXiv:2601.08287. — [arXiv:2601.08287](https://arxiv.org/abs/2601.08287) find
  18. Matthew Honnibal, Ines Montani, Sofie Van Lan-deghem, Adriane Boyd, and 1 others. 2020. spacy: Industrial-strength natural language processing in python. find
  19. He Hu, Yucheng Zhou, Juzheng Si, Qianning Wang, Hengheng Zhang, Fuji Ren, Fei Ma, Laizhong Cui, and Qi Tian. 2025. Beyond empathy: Integrating diagnostic and therapeutic reasoning with large language models for mental health counseling. arXiv preprint arXiv:2505.15715. — [arXiv:2505.15715](https://arxiv.org/abs/2505.15715) find
  20. Clayton Hutto and Eric Gilbert. 2014. Vader: A parsimonious rule-based model for sentiment analysis of social media text. In Proceedings of the international AAAI conference on web and social media, volume 8, pages 216–225. find
  21. Shaoxiong Ji, Tianlin Zhang, Luna Ansari, Jie Fu, Prayag Tiwari, and Erik Cambria. 2022. Mentalbert: Publicly available pretrained language models for mental healthcare. In proceedings of the thirteenth language resources and evaluation conference, pages 7184–7190. find
  22. Roman Kotov, David C Cicero, Christopher C Conway, Colin G DeYoung, Alexandre Dombrovski, Nicholas R Eaton, Michael B First, Miriam K Forbes, Steven E Hyman, Katherine G Jonas, and 1 others. 2022. The hierarchical taxonomy of psychopathology (hitop) in psychiatric practice and research. Psychologica… find
  23. Saif M Mohammad and Peter D Turney. 2013. Nrc emotion lexicon. National Research Council, Canada, 2:234. find
  24. Natalia B Mota, Nivaldo AP Vasconcelos, Nathalia Lemos, Ana C Pieretti, Osame Kinouchi, Guillermo A Cecchi, Mauro Copelli, and Sidarta Ribeiro. 2012. Speech graphs provide a quantitative measure of thought disorder in psychosis. PloS one, 7(4):e34928. find
  25. Niklas Muennighoff, Nouamane Tazi, Loïc Magne, and Nils Reimers. 2023. Mteb: Massive text embedding benchmark. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 2014–2037.
  26. Usman Naseem, Adam G Dunn, Jinman Kim, and Matloob Khushi. 2022. Early identification of depression severity levels on reddit using ordinal classification. In Proceedings of the ACM web conference 2022, pages 2563–2572. find
  27. Debora Nozza, Claudia Volpetti, and Elisabetta Fersini. 2019. Unintended bias in misogyny detection. In Ieee/wic/acm international conference on web intelligence, pages 149–155. find
  28. Charles O’Neill, Christine Ye, Kartheik Iyer, and John F. Wu. 2024. Disentangling dense embeddings with sparse autoencoders. Preprint, arXiv:2408.00657. — [arXiv:2408.00657](https://arxiv.org/abs/2408.00657) find
  29. Javier Parapar, Patricia Martín-Rodilla, David E Losada, and Fabio Crestani. 2021. erisk 2021: pathological gambling, self-harm and depression challenges. In European Conference on Information Retrieval, pages 650–656. Springer. find
  30. Anxo Pérez, Marcos Fernández-Pichel, Javier Parapar, and David E Losada. 2025. Depresym: A depression symptom annotated corpus and the role of large language models as assessors of psychological markers: A. pérez et al. Language Resources and Evaluation, 59(3):2737–2762. find
  31. Flor Miriam Plaza-del Arco, Debora Nozza, and Dirk Hovy. 2023. Respectful or toxic? using zero-shot learning with language models to detect hate speech. In The 7th workshop on online abuse and harms (woah), pages 60–68. find
  32. Federico Ravenda, Seyed Ali Bahrainian, Daniele Montagnani, Antonietta Mira, and Andrea Raballo. 2026. Personalitydbench: A dataset for personality disorders – from modeling to controlled generation. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics. Accepted…
  33. Federico Ravenda, Seyed Ali Bahrainian, Andrea Raballo, Antonietta Mira, and Noriko Kando. 2025. Are llms effective psychological assessors? leveraging adaptive rag for interpretable mental health screening through psychometric practice. In Proceedings of the 63rd Annual Meeting of the Association f… find
  34. Philip J Stone, Dexter C Dunphy, and Marshall S Smith. 1966. The general inquirer: A computer approach to content analysis. find
  35. Yla R Tausczik and James W Pennebaker. 2010. The psychological meaning of words: Liwc and computerized text analysis methods. Journal of language and social psychology, 29(1):24–54. find
  36. Kimberly A Van Orden, Tracy K Witte, Kelly C Cukrow-icz, Scott R Braithwaite, Edward A Selby, and Thomas E Joiner Jr. 2010. The interpersonal theory of suicide. Psychological review, 117(2):575. find
  37. Mehmet Can Yavuz. 2020. Analyses of character emotions in dramatic works by using emolex unigrams. In Proceedings of the Seventh Italian Conference on Computational Linguistics (CLiC-it 2020), pages 472–477. find
  38. Johannes Zimmermann, Timo Brockmeyer, Matthias Hunn, Henning Schauenburg, and Markus Wolf. 2017. First-person pronoun use in spoken language as a predictor of future depressive symptoms: Preliminary evidence from a clinical sample of depressed patients. Clinical psychology & psychotherapy, 24(2):384… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ravenda-2026-tony-open-source-toolkit-nlp/