Previous Slide | Background | Next Slide |
There are well-known gender differences in modern language use in less formal settings: conversations, e-mail, and letters.
Palander-Collin notes that female 17th century, English letters are more "involved, interactive, and personal" as well as more hedged ("I think").
Recent work (unpublished? by Koppel, Argamon, Shimoni, and Fine) on the British National Corpus demonstrates the ability to classify texts by gender of author using simple lexical and syntactic features.
After training, system correctly identifies gender of author 70% for non-fiction samples and more than 80% of samples of fiction.
Previous Slide | Olsen: Écriture Féminine...: Slide 3 | Next Slide |