Exploiting Lexical Dependencies from Large-Scale Data for Better Shift-Reduce Constituency Parsing

5810

Knowledge Triples