Experimenting with Transformer-based and Large Language Models for Classifying Effects of Outdoor Spaces on Social Anxiety in Social Media Data

AbstractSocial Anxiety Disorder (SAD) is a common condition, affecting a significant portion of the population. While research suggests spending time in nature can alleviate anxiety, the specific impact on SAD remains unclear. This study explores the relationship between discussions of outdoor spaces and social anxiety on social media. We leverage transformer-based and large language models (LLMs) to analyze a social media dataset focused on SAD. We developed three methods for the task of predicting the effects of outdoor spaces on SAD in social media. A two-stage pipeline classifier achieved the best performance of our submissions with results exceeding baseline performance.