Speech and Text Based Students' Mental Health Detection System Using Natural Language Processing with Explainable Ai

View at publisher Summarise with VIRAT
Authors: Janidu Himansa Welarathna, Pubudu Nallaperunma, Gayana Fernando
Cited by: 0
References: 15

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Mental health challenges among students have emerged as a pressing global issue, with students experiencing serious psychological distress and concentration difficulties due to stress. A significant problem is that many students cannot identify their mental health struggles or pinpoint what causes their academic pressure, resulting in unnoticed difficulties that escalate gradually. Existing evaluation approaches do not effectively detect the underlying factors behind academic pressure or offer transparent reasoning suitable for clinical applications. This research presents a dual-modality system that combines speech emotion recognition with text-based mental health classification using Natural Language Processing (NLP) and Explainable AI (XAI) techniques. The system analyzes emotional patterns in students' voices to identify eight emotions while simultaneously processing spoken content to classify mental health concerns into academic, non-academic, or mixed categories. Seven machine learning (ML) algorithms and one deep learning (DL) model were evaluated for emotion recognition, with Random Forest (RF) achieving 83.66% accuracy. For mental health classification, a fine-tuned BERT model achieved 98.91 % accuracy. XAI techniques including LIME and SHAP provide transparent interpretations by highlighting which words support academic versus non-academic classifications, making the system suitable as a clinical decision support tool. The system operates effectively in uncontrolled environments with language-independent audio features, providing clinicians with a dependable supplementary assessment resource that preserves their professional judgment authority.

Fields of Study & Subjects

Emotion and Mood Recognition Mental Health via Writing Sentiment Analysis and Opinion Mining

Publication details

Journal / source
International Conference on Advanced Research in Computing (ICARC)
Pages
1–6
Publication date
18 February 2026
Language
EN