A Distant Reading of Gender Bias in Dutch Literary Prizes

The Dutch literary scene has been criticized by authors for a lack of diversity and gender inequality. The two most important prizes, the Boekenbon
Literatuurprijs and the Libris Literatuur Prijs, show this gender inequality,
as about 80% of the nominated books were written by men, despite an
equal author gender distribution among published literary books in the
Netherlands. Given the over-representation of men in Dutch literary nominations, this inequality may be reflected in the word use of the authors, as
people tend to use similar language as their peers. Therefore, this paper
investigates whether it is possible to identify author gender inequality in
Dutch literary prizes using distant reading techniques: text classification,
topic modeling, and stylometry.
We collect a corpus of 300 literary books, divided into three categories:
nominated (Nom), not nominated books written by a nominated author
(NomAut), and books written by an author who has never been nominated
(NotNom). A classification model trained to predict the category of a book
reaches a cross-validated accuracy of 58.7%, surpassing the majority baseline (34%). Thus, nominated and not nominated books have distinctive
textual features, which supports the view that literary quality is associated
with particular formal features such as word usage. However, this word usage seems to be further removed from women writers, as the classification
of books written by women consistently shows the lowest performance.
The analysis of topics in the corpus suggest that the relation between nominated and not nominated books and author gender highly depends on the
topic which is investigated. The difference in writing style of nominated
and not nominated books cannot be clearly defined, but the results do
suggest that the writing style of Harry Mulisch and Herman Koch may
have influenced the writing styles of books nominated for literary prizes.