Introducing state of the art text classification with universal language models