Loading…
Title

Towards effective instance learning for neural dialogue ...

Description
By h cai · 2020 · cited by 19 — in this paper, we propose a data manipulation frame- work to proactively reshape the ... of data manipula- tion network and wi ∈ r is the importance weight.
/0
Full Text
Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6334–6343July 5 - 10, 2020. c©2020 Association for Computational Linguistics6334Data Manipulation: Towards Effective Instance Learningfor Neural Dialogue Generation via Learning to Augment and ReweightHengyi Cai†,§∗, Hongshen Chen‡Yonghao Song†, Cheng Zhang†, Xiaofang Zhao† and Dawei Yin†††Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China§University of Chinese Academy of Sciences, Beijing, China‡Data Science Lab, JD.com, China††Baidu Inc., [email protected], [email protected],{songyonghao, zhangcheng, zhaoxf}@ict.ac.cn, [email protected] state-of-the-art neural dialogue mod-els learn from human conversations followingthe data-driven paradigm. As such, a reliabletraining corpus is the crux of building a robustand well-behaved dialogue model. However,due to the open-ended nature of human con-versations, the quality of user-generated train-ing ...