Instead of relying only on document-to-question interaction or discrete features as in prior work, our model attends to relevant external knowledge and combines ...
確定! 回上一頁