A question-answer dataset and benchmark for testing machine reading.
SQuAD is a reading quiz with a highlighter. The answer must be in the paragraph, not from your wild guess.
It tests reading QA. It checks if a model can find the right answer in the text.
QA
SQuAD is one of the classic reading benchmarks for QA.
NLU
SQuAD uses reading questions to test if a model understands text.
BERT
BERT once made a big score jump on SQuAD.