We present a methodology that combines different techniques borrowed from computer science and the humanities in order to analyze character networks in a corpus of 1,800 British and American novels dating from 1849-1899 from a statistical standpoint. This methodology consists of three stages: preprocessing, locating characters, and associating characters. Using this technique, interactive visualizations of character networks are also produced. Franco Moretti, Alexander G. Sack, G. M. Park, and other humanists have presented different approaches to network analysis. While these studies increase our understanding of character networks, we noticed these analyses tend to focus on smaller corpora or individual novels. Using their studies as our starting point, we try to understand characters at a macroanalytical level. After testing our data and drawing some conclusions regarding authors’ gender and authors’ nationalities at a broad level, we have created a reliable methodology that can produce character data and be applied to future in-depth studies of character networks.
Full abstract available here