TY - GEN
T1 - Coarse-to-fine question answering for long documents
AU - Choi, Eunsol
AU - Hewlett, Daniel
AU - Uszkoreit, Jakob
AU - Polosukhin, Illia
AU - Lacoste, Alexandre
AU - Berant, Jonathan
N1 - Publisher Copyright:
© 2017 Association for Computational Linguistics.
PY - 2017
Y1 - 2017
N2 - We present a framework for question answering that can efficiently scale to longer documents while maintaining or even improving performance of state-of-the-art models. While most successful approaches for reading comprehension rely on recurrent neural networks (RNNs), running them over long documents is prohibitively slow because it is difficult to parallelize over sequences. Inspired by how people first skim the document, identify relevant parts, and carefully read these parts to produce an answer, we combine a coarse, fast model for selecting relevant sentences and a more expensive RNN for producing the answer from those sentences. We treat sentence selection as a latent variable trained jointly from the answer only using reinforcement learning. Experiments demonstrate the state of the art performance on a challenging subset of the WIKIREADING dataset (Hewlett et al., 2016) and on a new dataset, while speeding up the model by 3.5x-6.7x.
AB - We present a framework for question answering that can efficiently scale to longer documents while maintaining or even improving performance of state-of-the-art models. While most successful approaches for reading comprehension rely on recurrent neural networks (RNNs), running them over long documents is prohibitively slow because it is difficult to parallelize over sequences. Inspired by how people first skim the document, identify relevant parts, and carefully read these parts to produce an answer, we combine a coarse, fast model for selecting relevant sentences and a more expensive RNN for producing the answer from those sentences. We treat sentence selection as a latent variable trained jointly from the answer only using reinforcement learning. Experiments demonstrate the state of the art performance on a challenging subset of the WIKIREADING dataset (Hewlett et al., 2016) and on a new dataset, while speeding up the model by 3.5x-6.7x.
UR - http://www.scopus.com/inward/record.url?scp=85040945769&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85040945769&partnerID=8YFLogxK
U2 - 10.18653/v1/P17-1020
DO - 10.18653/v1/P17-1020
M3 - Conference contribution
AN - SCOPUS:85040945769
T3 - ACL 2017 - 55th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference (Long Papers)
SP - 209
EP - 220
BT - ACL 2017 - 55th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference (Long Papers)
PB - Association for Computational Linguistics (ACL)
T2 - 55th Annual Meeting of the Association for Computational Linguistics, ACL 2017
Y2 - 30 July 2017 through 4 August 2017
ER -