Evaluating shallow and deep learning strategies for the 2018 n2c2 shared task on clinical text classification