Depression Detection Using a Hybrid CNN-BiGRU-BiLSTM Model With BERT Embedding and Data Augmentation

Depression is a mood disorder that causes a persistent feeling of sadness and loss of interest. Also called major depressive disorder or clinical depression, it affects how you feel, think, and behave and can lead to a variety of emotional and physical problems. Automatic detection, as in a clinical interview can be derived from various modalities which includes video, audio, and text. Among them, detection from text data has emerged as a crucial task in mental health monitoring. In this study, we propose a hybrid deep learning model combining Convolutional Neural Networks (CNN), Bidirectional Gated Recurrent Units (BiGRU), and Bidirectional Long Short-Term Memory (BiLSTM) layers to enhance depression detection accuracy. Using BERT embedding for robust feature extraction, the proposed model processes social media or clinical text data to identify signs of depression. Additionally, we incorporate a novel data augmentation technique using synonym replacement to address data imbalance and improve generalization. Evaluations are conducted using key performance metrics, including accuracy, precision, recall, and F1 score. With an accuracy of 86.2 %, the results demonstrate that the combined CNN-BiGRU-BiLSTM architecture, alongside BERT embedding and augmented data, significantly improves classification performance compared to traditional models. This approach shows promise in contributing to more effective automatic depression detection systems.