ABSTRACT
Mental health is one of the pressing issues during the COVID-19 pandemic. Psychological distress can be caused directly by the pandemic itself, such as fear of contracting the disease, or by stress from losing jobs due to the disruption of economic activities. In addition, many government measures such as lockdown, unemployment aids, subsidies, or vaccination policy also affect population mood, sentiments, and mental health. This paper utilizes deep-learning-based techniques to extract sentiment, mood, and psychological signals from social media messages and use such aggregate signals to trace population-level mental health. To validate the accuracy of our proposed methods, we cross-check our results with the actual mental illness cases reported by Thailand’s Department of Mental Health and found a high correlation between the predicted mental health signals and the actual mental illness cases. Finally, we discuss potential applications that could be implemented using our proposed methods as building blocks. © 2021, Springer Nature Switzerland AG.