This project tries to parse out all entities in a novel and figure out, who is talking to who how often and what about. The resulting data is visualized interactively, so one can play with it and hopefully some interesting observations can be made.
The "A Song of Ice and Fire" series is a particularly interesting subject, due to the immense number of characters and the complex network of families, affiliations and politics.
The analysis is not perfect. Sometimes it isn't so easy to decide who is talking and to whom. The project is a work in progress. That being said, the vast majority of dialogues should be parsed correctly, even if the odd sentence is misattributed.
It's also quite strange, that according to the algorithm, Hodor spoke over a hundred different words. ;)
The source code can be found on github.