I am currently working on a question answering system. I create a synthetic dataset that contains multiple words in the answers. But, the answers are not a span of the given
