Extraction of Character Networks from TV Series Data

The analysis of character networks has been employed mainly in literary works, though they do not probably represent the majority of fictional products consumed nowadays since the attention of the audience at large has moved to audiovisual content (e.g., movies and TV series). This kind of analysis allows us to reveal narrative and structural properties. In this paper, we report the results of an extensive character network analysis carried out over all the seasons of the blockbuster series House, M.D. for a total of 177 episodes. We compute size, centrality, and connectivity metrics to analyse the interactions among the characters, but also to guess screenwriting choices that are not readily apparent. In particular, we show that: a) the choice of the relationship to describe through the network actually allows for different viewpoints; b) fixed characters exhibit significant differences in their relevance; c) the relevance assigned to secondary characters (i.e., other than House) has decreased over time.