Analyzing student mental health with RoBERTa-Large: a sentiment analysis and data analytics approach

Area of Science:

• Artificial Intelligence
• Psychology
• Natural Language Processing

Background:

• Student mental health is crucial for academic success and overall wellbeing.
• Traditional mental health monitoring methods are time-consuming and prone to bias.
• AI, specifically NLP and sentiment analysis, offers novel approaches to analyze textual data for mental health insights.

Purpose of the Study:

• To investigate student mental health using sentiment analysis with advanced deep learning models.
• To evaluate the effectiveness of Large Language Models (LLMs) in analyzing student mental health data.
• To identify patterns in student mental health influenced by academic and physical activities.

Main Methods:

• Sentiment analysis was performed using state-of-the-art Large Language Models (LLMs).
• Models evaluated included RoBERTa (a robustly optimized BERT approach), RoBERTa-Large, and ELECTRA.
• A multi-classification task with seven distinct mental health status labels was utilized.

Main Results:

• RoBERTa-Large achieved the highest accuracy at 97% in predicting student mental health status.
• ELECTRA demonstrated strong performance with 91% accuracy.
• The study highlights the efficacy of LLM-based sentiment analysis for mental health assessment.

Conclusions:

• LLM-based sentiment analysis is a powerful tool for monitoring student mental health.
• Advanced deep learning models can effectively identify complex emotional expressions and sentiments in textual data.
• This AI-driven approach shows significant potential for understanding the impact of academic and physical activities on student wellbeing.