Probabilistic Textual Time Series Depression Detection

2025-11-06

AI Summary

Existing depression severity prediction models lack uncertainty quantification and temporal dynamic modeling, limiting their interpretability and clinical reliability. To address this, we propose the first framework integrating sequential text modeling with probabilistic output: it employs bidirectional LSTMs and self-attention to extract features from multi-turn interview sequences, incorporates residual connections to enhance representational capacity, and adopts Gaussian or Student-t distributional output heads to yield well-calibrated prediction intervals. The model is trained end-to-end using negative log-likelihood loss and supports both sequence-to-scalar and sequence-to-sequence prediction. On the E-DAIC and DAIC-WOZ benchmarks, it achieves state-of-the-art performance for text-based depression assessment (E-DAIC MAE = 3.85), with significantly improved interval calibration. Ablation studies confirm the critical contributions of temporal modeling and uncertainty estimation. This work establishes a new paradigm for clinical decision support—balancing accuracy, robustness, and interpretability.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Accurate and interpretable predictions of depression severity are essential for clinical decision support, yet existing models often lack uncertainty estimates and temporal modeling. We propose PTTSD, a Probabilistic Textual Time Series Depression Detection framework that predicts PHQ-8 scores from utterance-level clinical interviews while modeling uncertainty over time. PTTSD includes sequence-to-sequence and sequence-to-one variants, both combining bidirectional LSTMs, self-attention, and residual connections with Gaussian or Student-t output heads trained via negative log-likelihood. Evaluated on E-DAIC and DAIC-WOZ, PTTSD achieves state-of-the-art performance among text-only systems (e.g., MAE = 3.85 on E-DAIC, 3.55 on DAIC) and produces well-calibrated prediction intervals. Ablations confirm the value of attention and probabilistic modeling, while comparisons with MentalBERT establish generality. A three-part calibration analysis and qualitative case studies further highlight the interpretability and clinical relevance of uncertainty-aware forecasting.

Problem

• Predicts depression severity from clinical interview texts
• Models temporal uncertainty in depression detection
• Provides interpretable uncertainty estimates for clinical decisions

Innovation

• Probabilistic framework combining LSTMs with attention
• Gaussian and Student-t output heads for uncertainty
• Sequence-to-sequence and sequence-to-one temporal modeling