Suicide Risk Assessment Using Multimodal Speech Features: A Study on the SW1 Challenge Dataset

TL;DR

This study explores a multimodal speech-based approach for adolescent suicide risk assessment, combining transcription, linguistic, and acoustic features with various fusion strategies, achieving 69% accuracy but facing generalization challenges.

Contribution

It introduces a multimodal fusion framework using advanced embeddings and compares different strategies, highlighting the importance of fusion mechanisms for suicide risk classification.

Findings

Weighted attention fusion achieved best accuracy (69%)

Generalization gap observed between development and test sets

Refinement of embeddings and fusion methods is crucial

Abstract

The 1st SpeechWellness Challenge conveys the need for speech-based suicide risk assessment in adolescents. This study investigates a multimodal approach for this challenge, integrating automatic transcription with WhisperX, linguistic embeddings from Chinese RoBERTa, and audio embeddings from WavLM. Additionally, handcrafted acoustic features -- including MFCCs, spectral contrast, and pitch-related statistics -- were incorporated. We explored three fusion strategies: early concatenation, modality-specific processing, and weighted attention with mixup regularization. Results show that weighted attention provided the best generalization, achieving 69% accuracy on the development set, though a performance gap between development and test sets highlights generalization challenges. Our findings, strictly tied to the MINI-KID framework, emphasize the importance of refining embedding representations and fusion mechanisms to enhance classification reliability.