Teaching machines to read natural language documents remains an elusive
challenge. machine reading systems can be tested on their ability to answer
questions posed on the contents of documents that they have seen, but until now
large scale training and test datasets have been missing f