Towards Domain-Generalizable Paraphrase Identification by Avoiding the Shortcut Learning

Xin Shen, Wai Lam


Abstract
In this paper, we investigate the Domain Generalization (DG) problem for supervised Paraphrase Identification (PI). We observe that the performance of existing PI models deteriorates dramatically when tested in an out-of-distribution (OOD) domain. We conjecture that it is caused by shortcut learning, i.e., these models tend to utilize the cue words that are unique for a particular dataset or domain. To alleviate this issue and enhance the DG ability, we propose a PI framework based on Optimal Transport (OT). Our method forces the network to learn the necessary features for all the words in the input, which alleviates the shortcut learning problem. Experimental results show that our method improves the DG ability for the PI models.
Anthology ID:
2021.ranlp-1.148
Volume:
Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021)
Month:
September
Year:
2021
Address:
Held Online
Editors:
Ruslan Mitkov, Galia Angelova
Venue:
RANLP
SIG:
Publisher:
INCOMA Ltd.
Note:
Pages:
1318–1325
Language:
URL:
https://aclanthology.org/2021.ranlp-1.148
DOI:
Bibkey:
Cite (ACL):
Xin Shen and Wai Lam. 2021. Towards Domain-Generalizable Paraphrase Identification by Avoiding the Shortcut Learning. In Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021), pages 1318–1325, Held Online. INCOMA Ltd..
Cite (Informal):
Towards Domain-Generalizable Paraphrase Identification by Avoiding the Shortcut Learning (Shen & Lam, RANLP 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.ranlp-1.148.pdf
Data
PARADE