My simple 2015 Giller Prize predictor.
predict.py is well-documented and reading through the main method will show you how my method works. Running predict.py performs a type of grid-search through model parameters and spits out the prediction for both the 2014 and 2015 Giller Prize contests.
My model predicts Russell Smith's Confidence as the winner, but if you consider a 4% margin of error, it's really a 5-way tie between Michael Christie's If I Fall, If I Die, Marina Endicott's Close to Hugh, Clifford Jackman's The Winter Family, and Anakana Schofield's Martin John. (There are 12 nominations in total so this doesn't quite bode well.)
UPDATE: Andre Alexis' Fifteen Dogs was announced as the winner of the 2015 Giller Prize and surprisingly, my model predicted this novel as winning with the least confidence. So if you invert my model, you have yourself a Giller prize predictor! Jokes, it just entails I have alot more work to do for the 2016 Giller Prize.
Unfortunately the data, i.e. the set of training, validation, and test novel texts, is not included in the repository and cannot be released for copyright reasons. Regardless, hopefully this repository serves useful for your own computational predictors for literature prize winners.
MIT