Utilizing BERT and TextCNN for Detecting Depression within Social Media Text Comments

Depression is a common symptom associated with various mental health conditions. Social media platforms have become a space for people to openly share their thoughts and feelings. These online expressions may have signs of depression. As such, the ability to detect depression in people’s social media content is a crucial step towards addressing and preventing mental health issues. Deep learning techniques have emerged as effective approaches for analyzing large textual datasets, such as social media posts. This paper examines two well-established deep learning (DL) models, namely BERT and TextCNN in detecting depressive symptoms from text. The findings suggest that both the BERT and TextCNN models exhibit the capability to identify signs of depression within text-based content. While the TextCNN model was found to have a faster training time, the BERT model demonstrated a higher overall accuracy rate in detecting depressive symptoms. These results highlight the potential of deploying both models and natural language processing techniques to gain insights into depressive symptoms from people’s online expressions.