Assessing Sentiment Segregation in Urban Communities

In this work, we attempt to examine the relationship between the social antecedents of urban neighborhoods and the citizens' everyday sentiment expression left in social media. Using Twitter users' geocoded messages posted within neighborhoods in the city of Pittsburgh, we first construct sentiment profiles for each neighborhood. We identify neighborhoods with relatively stable sentiment profiles and analyze the correlations between their sentiment orientations and the neighborhoods' demographic attributes including age, public safety, education, and ethnicity. The first order correlations show an interesting association between these neighborhood attributes and particular types of sentiments. We further group neighborhoods with similar demographic characteristics and observe that between two demographic groups, sentiments diverge in several sentiment categories including joy and disgust. This paper presents empirical evidence of sentiment segregation corresponding to the neighborhood contexts, which has implications for monitoring public attitudes and community integration.