Interpretable Natural Language Processing with Applications to Information Extraction