A Dataset and Strong Baselines for Classification of Czech News Texts