Effect of social capital on emotion, language style and latent topics in online depression community

Social capital is linked to mental illness. It has been proposed that higher social capital is associated with better mental well-being in both individuals and groups in offline setting. However, in online settings, the association between online social capital and mental health conditions has not yet been explored. Social media offer us a rich opportunity to determine the link between social capital and aspects of mental wellbeing. In this paper, we examine social capital based on levels of social connectivity of bloggers can be connected to aspects of depression in individuals and online depression community. We explore apparent properties of textual contents, including expressed emotions, language styles and latent topics, of a large corpus of blog posts, to analyze the aspect of social capital in the community. Using data collected from online Livejoumal depression community, we apply both statistical tests and machine learning approaches to examine how predictive factors vary between low and high social capital groups. Significant differences are found between low and high social capital groups when characterized by a set of latent topics, language features derived from blog posts, suggesting discriminative features, proved to be useful in the classification task. This shows that linguistic styles are better predictors than latent topics as features. The findings indicate the potential of using social media as a sensor for monitoring mental well-being in online settings.