Abstract—This paper presents the results of constraint-based automatic question generation for paragraphs from privacy policy documents. Existing work on question generation uses sequence-to-sequence and transformer-based approaches. This work introduces constraints to sequence-to-sequence and transformer based T5 model. The notion behind this work is that providing the deep learning models with additional background domain information can aid the system in learning useful patterns. This work presents three kinds of constraints – logical, empirical, and data-based constraint. The constraints are incorporated in the deep learning models by introducing additional penalty or reward terms in the loss function. Automatic evaluation results show that our approach significantly outperforms the state-of-the-art models.
Index Terms—Question generation, constraints, privacy policy, transformer, Sequence-to-Sequence model.
Deepti Lamba was a Ph.D. student at Kansas State University, Manhattan, KS 66506 USA (e-mail: dlamba@ksu.edu).
William H. Hsu is with Kansas State University, Manhattan, KS 66506 USA (e-mail: bhsu@ksu.edu).
[PDF]
Cite: Deepti Lamba and William H. Hsu, "Constraint-Based Neural Question Generation Using Sequence-to-Sequence and Transformer Models for Privacy Policy Documents," International Journal of Knowledge Engineering vol. 7, no. 2, pp. 14-20, 2021.
Copyright © 2021 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (
CC BY 4.0).
