%0 Generic %1 noauthororeditor %A Yang, Zhilin %A Qi, Peng %A Zhang, Saizheng %A Bengio, Yoshua %A Cohen, William %A Salakhutdinov, Ruslan %A Manning, Christopher D. %C Brussels, Belgium %D 2018 %I Association for Computational Linguistics %K qa %P 2369–2380 %R 10.18653/v1/D18-1259 %T HOTPOTQA: A Dataset for Diverse, Explainable Multi-hop Question Answering %U https://www.aclweb.org/anthology/D18-1259.pdf %V Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing %X Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. We introduce HOTPOTQA, a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) we provide sentence-level supporting facts required for reasoning, allowing QA systems to reason with strong supervision and explain the predictions; (4) we offer a new type of factoid comparison questions to test QA systems’ ability to extract relevant facts and perform necessary comparison. We show that HOTPOTQA is challenging for the latest QA systems, and the supporting facts enable models to improve performance and make explainable predictions.

@conference{noauthororeditor, abstract = {Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. We introduce HOTPOTQA, a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) we provide sentence-level supporting facts required for reasoning, allowing QA systems to reason with strong supervision and explain the predictions; (4) we offer a new type of factoid comparison questions to test QA systems’ ability to extract relevant facts and perform necessary comparison. We show that HOTPOTQA is challenging for the latest QA systems, and the supporting facts enable models to improve performance and make explainable predictions.}, added-at = {2019-12-07T11:26:23.000+0100}, address = {Brussels, Belgium}, author = {Yang, Zhilin and Qi, Peng and Zhang, Saizheng and Bengio, Yoshua and Cohen, William and Salakhutdinov, Ruslan and Manning, Christopher D.}, biburl = {https://www.bibsonomy.org/bibtex/2b749abd20ce9f1c5eb8ba8e0ea9b2b57/cyn7hia}, description = {HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering}, doi = {10.18653/v1/D18-1259}, interhash = {47f9e8025fb748932e2f0bc03b97de94}, intrahash = {b749abd20ce9f1c5eb8ba8e0ea9b2b57}, keywords = {qa}, month = {October-November}, pages = {2369–2380}, publisher = {Association for Computational Linguistics}, timestamp = {2019-12-09T08:42:51.000+0100}, title = {HOTPOTQA: A Dataset for Diverse, Explainable Multi-hop Question Answering}, url = {https://www.aclweb.org/anthology/D18-1259.pdf}, venue = {EMNLP}, volume = {Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing}, year = 2018 }