CANARD has been preprocessed by Voskarides et al. to train and evaluate their Query Resolution Term Classification
model (QuReTeC).
CANARD is a dataset for question-in-context rewriting that consists of questions each given in a dialog context
together with a context-independent rewriting of the question. The context of each question is the dialog utterences
that precede the question. CANARD can be used to evaluate question rewriting models that handle important linguistic
phenomena such as coreference and ellipsis resolution.
