1 paper · 1 filter
Jungo Kasai, Keisuke Sakaguchi, Ronan Le Bras +5
Natural language processing researchers have identified limitations of evaluation methodology for generation tasks, with new questions raised about the validity of automatic metric…