DEP-BERTCNN: A Weighted Deep Learning Model for Depression Detection in Online Forums

Anonymous

Abstract: Depression is a prevalent mental health disorder affecting a significant portion of the global population. With the rise of social media, online forums have become a popular platform for individuals to discuss their mental health concerns, and online forums have emerged as a valuable source of data for studying depression due to the vast amounts of user-generated content they contain. In this paper, we propose a weighted deep learning model named DEP-BERTCNN for depression detection in text in online forums. DEP-BERTCNN uses a combination of a pre-trained BERT language model, an attention model and convolutional neural network to classify forum users as either depressive of non-depressive. Also in DEP-BERTCNN, we computes weights, using the TFIDF method, based on linguistic depression indicators present in users' posts to enhance the performance of the depression detection model. These weights amplify the most informative aspects of posts indicative of depression. To the best of our knowledge, our work is the first to use input embeddings weighted with depression indicators in combination with an attention model for depression detection. The DEP-BERTCNN model was trained and evaluated on the large-scale Reddit Self-reported Depression Dataset (RSDD). Our results demonstrate that the proposed model outperform several baseline methods on the RSDD dataset, demonstrating the effectiveness of combining deep learning model with linguistic indicators associated with depression symptoms.

Paper Type: long

Research Area: NLP Applications

Contribution Types: NLP engineering experiment

Languages Studied: English

Consent To Share Submission Details: On behalf of all authors, we agree to the terms above to share our submission details.