Momentary Depressive Feeling Detection Using X (Formerly Twitter) Data: Contextual Language Approach.

Area of Science:

• Computational linguistics
• Mental health informatics
• Social media analytics

Background:

• Depression and momentary depressive feelings represent significant public health challenges.
• Early detection of these feelings is crucial for reducing societal burden and enhancing individual well-being.
• Social media platforms like X (formerly Twitter) offer vast data for understanding mental states and detecting depressive feelings.

Purpose of the Study:

• To automate the detection of momentary depressive feelings in social media posts.
• To evaluate the effectiveness of various machine learning and transfer learning models for this task.

Main Methods:

• Developed a lexicon of terms related to depression and momentary depressive feelings.
• Scraped and manually labeled social media posts using the lexicon.
• Assessed the performance of Bidirectional Encoder Representations From Transformers (BERT), A Lite BERT (ALBERT), Robustly Optimized BERT Approach (RoBERTa), Distilled BERT (DistilBERT), convolutional neural network (CNN), bidirectional long short-term memory (BiLSTM), and traditional machine learning (ML) algorithms.

Main Results:

• Binary classification outperformed multilabel classification in detecting depressive sentiments.
• Distilled BERT achieved the highest performance across multiple metrics (AUC, accuracy, sensitivity, specificity, precision, F1-score).
• Transfer learning algorithms, particularly DistilBERT, significantly outperformed traditional ML algorithms in contextual analysis and detection.

Conclusions:

• Contextual language approaches, especially transfer learning models, reliably automate early detection of momentary depressive feelings.
• These methods can inform the development of social media monitoring tools for identifying at-risk individuals.
• Proactive intervention through such tools can mitigate depression's societal impact and improve global public health outcomes.