This is a joint presentation. Graham Sack will present on character networks. David Mimno will present on the use of topic analysis to analyze character relationships. David Elson (Columbia University CS
Dept.) also contributed to this work.
We study social form in sixty 19th Century British novels. We begin by generating character co-occurrence networks for each novel using supervised named entity recognition. We compare network properties between novels using network analysis software. We then analyze topics associated with each relationship based on the language describing character interactions. This method exposes variation in network metrics (e.g., graph density and clustering coefficients) and topic prevalence, cross-tabulated and regressed against potential explanatory factors, including author, setting, point of view, and genre attributes.
Back to Networks and Network Analysis for the Humanities: Reunion Conference