WikiQA dataset for Answer Sentence Selection. The dataset contains 2 additional splits which are clean versions of the original development and test sets. clean versions contain only questions which have at least a positive and a negative answer candidate.
