You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We present NewsQA, a challenging machine comprehension dataset of over100,000 human-generated question-answer pairs. Crowdworkers supply questionsand answers based on a set of over 10,000 news articles from CNN, with answersconsisting of spans of text from the corresponding articles. We collect thisdataset through a four-stage process designed to solicit exploratory questionsthat require reasoning. A thorough analysis confirms that NewsQA demandsabilities beyond simple word matching and recognizing textual entailment. Wemeasure human performance on the dataset and compare it to several strongneural models. The performance gap between humans and machines (0.198 in F1)indicates that significant progress can be made on NewsQA through futureresearch. The dataset is freely available athttps://datasets.maluuba.com/NewsQA.
URL
Affiliations
Abstract
Translation (by gpt-3.5-turbo)
Summary (by gpt-3.5-turbo)
The text was updated successfully, but these errors were encountered: