Export Reviews, Discussions, Author Feedback and Meta-Reviews

This paper focuses on the reading comprehension task. It proposes to construct a large dataset using news stories and highlights from CNN and Daily Mail. Reading comprehension is formulated as a Cloze test, where a named entity in the news highlight is replaced by a placeholder. The system is expected to predict the missing entity (answer) based on the entire news document (document) and highlight (query). The paper explores several approaches to address the problem.