Machine learning models possess a fundamental limitation that often frustrates newcomers to natural language processing (NLP): they cannot read.