De-identification of clinical free text using natural language processing: A systematic review of current approaches