Parsing clinical text: how good are the state-of-the-art parsers?