Combining Lexical and Syntactic Features for Detecting Content-Dense Texts in News