Elamur Research⌕
☰
home / papers / 2026 / Probabilistic Depression Detection from Textual Time Series

Probabilistic Depression Detection from Textual Time Series

Paper · 2026 · Fabian Schmidt, Seyedehmoniba Ravan and Vladimir Vlassov

Summary

Text-based depression severity models usually output a single number with no sign of how reliable it is. That limits their use in clinical decision support. The authors present PTTSD, which treats an interview as a sequence of utterances and predicts a full probability distribution over the PHQ-8 score rather than a point value. Each utterance is embedded with a pretrained sentence encoder and passed through an LSTM with multi-head self-attention and residual connections. Separate heads then output the parameters of a Gaussian or Student's-t distribution, trained with negative log-likelihood. The seq-to-one variant scores the whole session. The seq-to-seq variant gives an estimate after every utterance prefix, which shows when confidence stabilises and which utterances shift the prediction. On DAIC-WOZ and E-DAIC, PTTSD matches or exceeds recent text-only systems without prompt engineering or handcrafted features. Its predicted uncertainty also closely tracks actual error. The practical value is that clinicians would receive both a severity estimate and a meaningful confidence signal.

Key points

  • Data: DAIC-WOZ (189 sessions; official split 107/35/56) and E-DAIC (275 sessions; split 163/56/56). E-DAIC audio was re-transcribed with WhisperX. PHQ-8 scores range from 0 to 24.
  • Embeddings: the primary encoder is all-MiniLM-L6-v2 (about 22M parameters). MentalBERT is tested as a domain-adapted alternative.
  • Accuracy: E-DAIC MAE = 3.85, RMSE = 4.52; DAIC MAE = 3.55, RMSE = 4.77. Accuracy drops on the original E-DAIC transcripts compared with the WhisperX versions, but calibration holds for both sources.
  • Calibration and analysis: predicted uncertainty correlates strongly with absolute error (r = 0.88, ECE = 0.022). Ablations support the attention mechanism and the probabilistic loss. Case studies of four participants illustrate four regimes, including an overconfident error.
  • Limitations: the model is text-only and the datasets are small. Interviews were run by a Wizard-of-Oz virtual agent ("Ellie"). Utterances are encoded without context-aware fine-tuning, and the system has no clinical utility evaluation.

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

🎓Uppsala UniversityUppsala · Sweden64 works · 13,164 citations🎓KTH Royal Institute of TechnologyStockholm · Sweden18 works · 788 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 worksSocial Psychology · 306 works

Cite

Cite as Fabian Schmidt, Seyedehmoniba Ravan and Vladimir Vlassov. 2026. Probabilistic Depression Detection from Textual Time Series. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1630

@inproceedings{schmidt2026probabilistic,
  title = {{Probabilistic Depression Detection from Textual Time Series}},
  author = {Fabian Schmidt, Seyedehmoniba Ravan and Vladimir Vlassov},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1630},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1630}
}

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. Tuka Al Hanai, Mohammad M Ghassemi, and James R Glass. 2018. Detecting depression with audio/text sequence modeling of interviews. In Interspeech, pages 1716–1720. find
  2. Sergio Burdisso, Esaú Villatoro-Tello, Srikanth Madikeri, and Petr Motlicek. 2023. Node-weighted graph convolutional network for depression detection in transcribed clinical interviews. In INTERSPEECH 2023, pages 3617–3621. find
  3. Zhuang Chen, Jiawen Deng, Jinfeng Zhou, Jincenzi Wu, Tieyun Qian, and Minlie Huang. 2024. Depression detection in clinical interviews with LLM-empowered structural element graph. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Hum… find
  4. DAIC-WOZ Project. 2019. Extended distress analysis interview corpus–wizard of oz (e-daic). Extended DAIC Database, downloadable via the DAIC-WOZ project website at dcapswoz.ict.usc.edu. find
  5. AVEC 2019 subset: 275 sessions (163 train, 56 dev, 56 test); includes audio, transcripts, visual and acoustic features; Accessed: 2025-01-30. find
  6. Mamadou Dia, Ghazaleh Khodabandelou, and Alice Othmani. 2024. Paying attention to uncertainty: A stochastic multimodal transformers for post-traumatic stress disorder detection using video. Computer Methods and Programs in Biomedicine, 257:108439. find
  7. Ming Fang, Siyu Peng, Yujia Liang, Chih-Cheng Hung, and Shuhua Liu. 2023a. A multimodal fusion model with multi-level attention mechanism for depression detection. Biomedical Signal Processing and Control, 82:104561. find
  8. Ming Fang, Siyu Peng, Yujia Liang, Chih-Cheng Hung, and Shuhua Liu. 2023b. A multimodal fusion model with multi-level attention mechanism for depression detection. Biomedical Signal Processing and Control, 82:104561. find
  9. Yuan Gong and Christian Poellabauer. 2017. Topic modeling based multi-modal depression detection. In Proceedings of the 7th Annual Workshop on Audio/Visual Emotion Challenge, AVEC ’17, page 69–76, New York, NY, USA. Association for Computing Machinery. find
  10. Jonathan Gratch, Ron Artstein, Gale Lucas, Giota Stratou, Stefan Scherer, Angela Nazarian, Rachel Wood, Jill Boberg, David DeVault, Stacy Marsella, David Traum, Skip Rizzo, and Louis-Philippe Morency. 2014. The distress analysis interview corpus of human and computer interviews. In Proceedings of th… find
  11. Yanrong Guo, Chenyang Zhu, Shijie Hao, and Richang Hong. 2022. A topic-attentive transformer-based model for multimodal depression detection. arXiv preprint arXiv:2206.13256. — [arXiv:2206.13256](https://arxiv.org/abs/2206.13256) find
  12. Shaoxiong Ji, Tianlin Zhang, Luna Ansari, Jie Fu, Prayag Tiwari, and Erik Cambria. 2022. Mental-BERT: Publicly available pretrained language models for mental healthcare. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 7184–7190, Marseille, France. European Langu… find
  13. Alex Kendall and Yarin Gal. 2017. What uncertainties do we need in bayesian deep learning for computer vision? In Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc.
  14. Kurt Kroenke, Tara W Strine, Robert L Spitzer, Janet BW Williams, Joyce T Berry, and Ali H Mokdad. 2009. The phq-8 as a measure of current depression in the general population. Journal of affective disorders, 114(1-3):163–173. find
  15. Yonghong Li and Xiuzhuang Zhou. 2025. Fair uncertainty quantification for depression prediction. arXiv preprint arXiv:2505.04931. — [arXiv:2505.04931](https://arxiv.org/abs/2505.04931) find
  16. Mariana Rodrigues Makiuchi, Tifani Warnita, Kuniaki Uto, and Koichi Shinoda. 2019. Multimodal fusion of bert-cnn and gated cnn representations for depression detection. In Proceedings of the 9th International on Audio/Visual Emotion Challenge and Workshop, AVEC ’19, page 55–63, New York, NY, USA. As… find
  17. Aishik Mandal, Dana Atzil-Slonim, Thamar Solorio, and Iryna Gurevych. 2025. Enhancing depression detection via question-wise modality fusion. In Proceedings of the 10th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2025), pages 44–61, Albuquerque, New Mexico. Association for… find
  18. Kaining Mao, Wei Zhang, Deborah Baofeng Wang, Ang Li, Rongqi Jiao, Yanhui Zhu, Bin Wu, Tiansheng Zheng, Lei Qian, Wei Lyu, Minjie Ye, and Jie Chen. 2023. Prediction of depression severity based on the prosodic and semantic features with bidirectional lstm and time distributed cnn. IEEE Transactions… find
  19. Niklas Muennighoff, Nouamane Tazi, Loic 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, Dubrovnik, Croatia. Association for Computational Linguistic…
  20. Meng Niu, Kai Chen, Qingcai Chen, and Lufeng Yang. 2021. Hcag: A hierarchical context-aware graph attention model for depression detection. In ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 4235–4239. find
  21. Mariia Nykoniuk, Oleh Basystiuk, Nataliya Shakhovska, and Nataliia Melnykova. 2025. Multimodal data fusion for depression detection approach. Computation, 13(1):9. find
  22. Syed Arbaaz Oureshi, Gaël Dias, Sriparna Saha, and Mohammed Hasanuzzaman. 2021. Gender-aware estimation of depression severity level in a multimodal setting. In 2021 International Joint Conference on Neural Networks (IJCNN), pages 1–8. find
  23. Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Te-jani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, and 2 others. 20… — PyTorch: An Imperative Style, High-Performance Deep Learning Library
  24. Syed Arbaaz Qureshi, Mohammed Hasanuzzaman, Sriparna Saha, and Gaël Dias. 2019a. The verbal and non verbal signals of depression–combining acoustics, text and visuals for estimating depression level. arXiv preprint arXiv:1904.07656. — [arXiv:1904.07656](https://arxiv.org/abs/1904.07656) find
  25. Syed Arbaaz Qureshi, Sriparna Saha, Mohammed Hasanuzzaman, and Gaël Dias. 2019b. Multitask representation learning for multimodal estimation of depression level. IEEE Intelligent Systems, 34(5):45– 52. find
  26. Anupama Ray, Siddharth Kumar, Rutvik Reddy, Prerana Mukherjee, and Ritu Garg. 2019. Multi-level attention network using text, audio and video for depression prediction. In Proceedings of the 9th International on Audio/Visual Emotion Challenge and Workshop, AVEC ’19, page 81–88, New York, NY, USA. As… find
  27. Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence embeddings using Siamese BERT-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3982–3992,…
  28. Morteza Rohanian, Julian Hough, and Matthew Purver. 2019. Detecting depression with word-level multimodal fusion. In INTERSPEECH 2019, pages 1443– 1447. find
  29. Misha Sadeghi, Bernhard Egger, Reza Agahi, Robert Richer, Klara Capito, Lydia Helene Rupp, Lena Schindler-Gmelch, Matthias Berking, and Bjoern M. Eskofier. 2023. Exploring the capabilities of a language model-only approach for depression detection in text data. In 2023 IEEE EMBS International Confer… find
  30. Misha Sadeghi, Robert Richer, Bernhard Egger, Lena Schindler-Gmelch, Lydia Helene Rupp, Farnaz Rahimi, Matthias Berking, and Bjoern M Eskofier. 2024. Harnessing multimodal approaches for depression detection using large language models and facial expressions. npj Mental Health Research, 3(1):66. find
  31. Maximilian Seitzer, Arash Tavakoli, Dimitrije Antic, and Georg Martius. 2022. On the pitfalls of heteroscedastic uncertainty estimation with probabilistic neural networks. In International Conference on Learning Representations. find
  32. Evgeny A. Stepanov, Stéphane Lathuilière, Shammur Absar Chowdhury, Arindam Ghosh, Radu-Laurenţiu Vieriu, Nicu Sebe, and Giuseppe Riccardi. 2018. Depression severity estimation from multiple modalities. In 2018 IEEE 20th International Conference on e-Health Networking, Applications and Services (Hea… find
  33. WHO. 2017. Depression and other common mental disorders: Global health estimates. Technical report, World Health Organization, Geneva. WHO/MSD/MER/2017.2. find
  34. WHO. 2022. World mental health report: Transforming mental health for all. Accessed: 2025-05-18. find
  35. James R. Williamson, Elizabeth Godoy, Miriam Cha, Adrianne Schwarzentruber, Pooya Khorrami, Youngjune Gwon, Hsiang-Tsung Kung, Charlie Dagli, and Thomas F. Quatieri. 2016. Detecting depression using vocal, facial and semantic communication cues. In Proceedings of the 6th International Workshop on Au… find
  36. Le Yang, Dongmei Jiang, Xiaohan Xia, Ercheng Pei, Meshia Cédric Oveneke, and Hichem Sahli. 2017. Multimodal measurement of depression using deep learning models. In Proceedings of the 7th Annual Workshop on Audio/Visual Emotion Challenge, AVEC ’17, page 53–59, New York, NY, USA. Association for Comp… find
  37. Jun Zhang and Yanrong Guo. 2024. Multilevel depression status detection based on fine-grained prompt learning. Pattern Recogn. Lett., 178(C):167–173. find
  38. Xu Zhang, Chenlong Li, Weisi Chen, Jiaxin Zheng, and Feihong Li. 2025. Optimizing depression detection in clinical doctor-patient interviews using a multi-instance learning framework. Scientific Reports, 15(1):6637. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/schmidt-2026-probabilistic-depression-detection-textual-time/