The application of a pre-trained language model to classify text into categories without explicit training on labeled examples of those categories.