Robust Speech and Natural Language Processing Models for Depression Screening

2020-12-05

IEEE Signal Processing in Medicine and Biology Symposium

AI Summary

To address the need for remote, contactless preliminary depression screening, this paper proposes a dual-modality transfer learning framework: separate deep models are built on acoustic features (MFCCs, prosody) and conversational text (BERT-based representations) for binary depression classification. The key contribution is the first validation of cross-speaker robustness on a large-scale, real-world human–machine dialogue dataset with over 10,000 annotated utterances—where test speakers are strictly disjoint from training speakers. Both models achieve AUC ≥ 0.80 on unseen speakers and diverse conversational scenarios, demonstrating strong generalization and practical deployability. This work establishes an interpretable, reproducible, and highly robust technical pathway for large-scale, population-level mental health screening.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Conversational AI/Dialog Systems
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Depression is a global health concern with a critical need for increased patient screening. Speech technology offers advantages for remote screening but must perform robustly across patients. We have described two deep learning models developed for this purpose. One model is based on acoustics; the other is based on natural language processing. Both models employ transfer learning. Data from a depression-labeled corpus in which 11,000 unique users interacted with a human-machine application using conversational speech is used. Results on binary depression classification have shown that both models perform at or above AUC=0.80 on unseen data with no speaker overlap. Performance is further analyzed as a function of test subset characteristics, finding that the models are generally robust over speaker and session variables. We conclude that models based on these approaches offer promise for generalized automated depression screening.

Problem

• Depression Detection
• Language Processing
• Speech Processing

Innovation

• Deep Learning
• Depression Detection
• Transfer Learning