a corpus containing 4.5K conversations from the Conversational Question-Answering dataset CoQA, for a total of 53K follow-up question-answer pairs, in which each original question was manually annotated with at least 2 and at most 3 out-of-context rewritings - View it on GitHub