A narrative review on human anxiety detection in social media using natural language processing

Abstract

Currently, with the boundless proliferation of social media platforms, it is undoubted that the world is in the middle of a significant phase of social media evolution. These virtually created user environments allow users to create, share, and interact with content. Most people who interact with social media make a pitch to share their mood, tension, feelings, and behaviour unhesitatingly with the community. Mostly, these data can be taken as a mirror that reflects the mental health status of a person, such as stress, anxiety, and suffering. With increased social media penetration, people tend to share their disturbances rather than suffer alone. Anxiety can be taken as an ordinary human emotion which prepares the human body for potentially vulnerable situations. It would be beneficial if there were any system that could predict if a person is anxious before going through a critical situation that would have to be clinically treated. This study is a narrative review of anxiety detection over the past ten years. Furthermore, this qualitative study discusses an overview of anxiety detection using Natural Language Processing (NLP) and key concepts highlighting emerging trends. We conclude that the existing state-of-the-art social media anxiety detection mechanisms can be outperformed using deep learning based on large pretrained language models. Furthermore, it can be safely stated that a specific pre-trained large language model for social media anxiety detection is a crucial necessity yet to be fulfilled.