Topical Classification and Divergence on Reddit

Online discussion forums are virtual spaces for communities of users to discuss various topics. As an online discussion progresses, there is a tendency for comments to diverge from the initial topic, often degenerating into off-topic shouting matches. We would like to determine how this divergence happens, and identify comments that trigger this topic drift. We approach this problem by first building a neural topic model, which we later use to analyze topic divergence at aggregate and individual thread levels.