Analyzing Latent Topics in Student Confessions Communities on Facebook

Abstract

Confessions pages have grown popular on social media sites such as Facebook and Twitter, particularly within college communities. Such pages allow users to anonymously submit confessions related to collegiate experience that are subsequently broadcast on a public forum. Because of the anonymous nature of disclosure, we believe that confessions pages are novel data sources from which to discover trends and issues in a collegiate community. Aggregating data from more than 20,000 entries posted to one such space, we analyze natural language characteristics of the originating community with LDA, pointwise mutual information and sentiment analysis. Using a Markov topic model, we identify the latent topics in our corpus and find that loneliness is a highly regular pattern. Our findings on student confession communities support previous sociological research, contextualizing student loneliness in the age of social networks.