Because of the various ways novelists inscribe character, it is far more difficult for a computer to identify the characters in a novel than the themes, and it is even more difficult for the machine to figure out what those characters are doing.
do their models take into account error in identifying characters?