A pre-trained language model used without further training to extract features or measurements from text.