TY - GEN
T1 - QG-Net
T2 - 5th Annual ACM Conference on Learning at Scale, L at S 2018
AU - Wang, Zichao
AU - Lan, Andrew S.
AU - Nie, Weili
AU - Waters, Andrew E.
AU - Grimaldi, Phillip J.
AU - Baraniuk, Richard G.
N1 - Funding Information:
This research was supported by the Arthur & Carlyse Ciocca Charitable Foundation, The Laura and John Arnold Foundation, John and Ann Doerr, IBM Research, and NSF grant DRL-1631556. Thanks to Bob Schloss for his insights on the work.
Publisher Copyright:
© 2017 Association for Computing Machinery. All rights reserved.
PY - 2018/6/26
Y1 - 2018/6/26
N2 - The ever growing amount of educational content renders it increasingly difficult to manually generate sufficient practice or quiz questions to accompany it. This paper introduces QG-Net, a recurrent neural network-based model specifically designed for automatically generating quiz questions from educational content such as textbooks. QG-Net, when trained on a publicly available, general-purpose question/answer dataset and without further fine-tuning, is capable of generating high quality questions from textbooks, where the content is significantly different from the training data. Indeed, QG-Net outperforms state-of-the-art neural network-based and rules-based systems for question generation, both when evaluated using standard benchmark datasets and when using human evaluators. QGNet also scales favorably to applications with large amounts of educational content, since its performance improves with the amount of training data.
AB - The ever growing amount of educational content renders it increasingly difficult to manually generate sufficient practice or quiz questions to accompany it. This paper introduces QG-Net, a recurrent neural network-based model specifically designed for automatically generating quiz questions from educational content such as textbooks. QG-Net, when trained on a publicly available, general-purpose question/answer dataset and without further fine-tuning, is capable of generating high quality questions from textbooks, where the content is significantly different from the training data. Indeed, QG-Net outperforms state-of-the-art neural network-based and rules-based systems for question generation, both when evaluated using standard benchmark datasets and when using human evaluators. QGNet also scales favorably to applications with large amounts of educational content, since its performance improves with the amount of training data.
UR - http://www.scopus.com/inward/record.url?scp=85051466066&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85051466066&partnerID=8YFLogxK
U2 - 10.1145/3231644.3231654
DO - 10.1145/3231644.3231654
M3 - Conference contribution
AN - SCOPUS:85051466066
T3 - Proceedings of the 5th Annual ACM Conference on Learning at Scale, L at S 2018
BT - Proceedings of the 5th Annual ACM Conference on Learning at Scale, L at S 2018
PB - Association for Computing Machinery, Inc
Y2 - 26 June 2018 through 28 June 2018
ER -