QED stands for Dataset for Explanations in Question Answering. It is a question-answering dataset annotated by experts and is based on a subset of the Google Natural Questions dataset.

QED is a linguistically principled framework for question-answering parsing. Given a question and a text, QED represents the parsing of the answer as a combination of discrete, understandable steps:

• Sentence selection: Find a sentence that contains an implicit question.
• Referential equality: Identify noun phrases in question and answer sentences that refer to the same thing.
• Predicate entailment: Once referential equality is extracted, it confirms that the predicate in the sentence contains the predicate in the question.