BERT-Based Text Classification Pipeline for Categorising Mental Health Statements
摘要
The automatic classification of mental health-related textual data is essential for enabling scalable and proactive psychological support. This study employs a fine-tuned BERT-base transformer model categorise statements into seven clinically relevant mental health conditions. The model was evaluated using a publicly available dataset from Kaggle and assessed with standard performance metrics, including accuracy, precision, recall, specificity, and F1-score. Achieving an overall accuracy of 90.00%, the model demonstrated strong generalisability across a linguistically diverse and imbalanced corpus. The highest classification performance was achieved in the Normal, Depression, and Suicidal categories, with satisfactory results obtained for less frequent classes. These findings highlight the effectiveness of transformer-based architectures in mental health text classification and support their potential application in digital screening and early intervention systems.