As part of my quest to benchmark text classification algorithms against Google’s new WikiReading dataset, I fed it into fastText today: