Explainable Depression Detection using Masked Hard Instance Mining

TL;DR

This paper introduces Masked Hard Instance Mining (MHIM), a novel method that enhances explainability and accuracy in text-based depression detection across multiple languages by strategically masking attention weights.

Contribution

The paper presents MHIM, a new approach that improves model interpretability and performance in depression detection by manipulating attention mechanisms.

Findings

MHIM improves prediction accuracy on Thai and English datasets.

MHIM enhances explainability metrics in depression detection models.

MHIM outperforms baseline models in both accuracy and interpretability.

Abstract

This paper addresses the critical need for improved explainability in text-based depression detection. While offering predictive outcomes, current solutions often overlook the understanding of model predictions which can hinder trust in the system. We propose the use of Masked Hard Instance Mining (MHIM) to enhance the explainability in the depression detection task. MHIM strategically masks attention weights within the model, compelling it to distribute attention across a wider range of salient features. We evaluate MHIM on two datasets representing distinct languages: Thai (Thai-Maywe) and English (DAIC-WOZ). Our results demonstrate that MHIM significantly improves performance in terms of both prediction accuracy and explainability metrics.