Character Identification in Children Stories

The automatic identification of direct and indirect discourses and the association of each “direct” utterance with its author are research topics that begin to be explored in Natural Language Processing. We developed the DID system that when applied to children stories starts by classifying the utterances that belong to the narrator (indirect discourse) and those belonging to the characters taking part in the story (direct discourse). Afterword, DID tries to associate each direct discourse utterance with the character(s) in the story. This automation can be advantageous, namely when it is necessary to tag the stories that should be handled by an automatic story teller.