Automated Depression Detection From Text and Audio: A Systematic Review.

Area of Science:

• Artificial Intelligence in Mental Health
• Computational Linguistics
• Machine Learning for Healthcare

Background:

• Depression is a widespread mental health issue requiring efficient diagnosis and treatment.
• Automated Depression Detection (ADD) systems offer scalable solutions using text and audio data.
• Current challenges include timely diagnosis and intervention for depression.

Purpose of the Study:

• To systematically review machine learning-based ADD systems using multimodal data.
• To analyze methodologies like data augmentation, multimodal fusion, and feature extraction.
• To identify current trends and future research directions in automated depression detection.

Main Methods:

• Systematic literature review of 65 studies (2018-2024).
• Focus on machine learning models utilizing multimodal (text and audio) data.
• Examination of data augmentation, fusion techniques, and feature extraction strategies.

Main Results:

• Identified key methodologies and state-of-the-art ADD systems.
• Highlighted the importance of culturally adaptable, high-quality datasets.
• Noted limitations in longitudinal data and real-world clinical integration.

Conclusions:

• Future ADD systems require interpretability, scalability, and robustness for clinical use.
• Emphasis on developing cross-cultural and clinically integrated depression detection tools.
• The review provides a comprehensive overview and identifies research gaps for advancing ADD.