Adversarial Training for Commonsense Inference

Lis Pereira, Xiaodong Liu, Fei Cheng, Masayuki Asahara, Ichiro Kobayashi


Abstract
We apply small perturbations to word embeddings and minimize the resultant adversarial risk to regularize the model. We exploit a novel combination of two different approaches to estimate these perturbations: 1) using the true label and 2) using the model prediction. Without relying on any human-crafted features, knowledge bases, or additional datasets other than the target datasets, our model boosts the fine-tuning performance of RoBERTa, achieving competitive results on multiple reading comprehension datasets that require commonsense inference.
Anthology ID:
2020.repl4nlp-1.8
Volume:
Proceedings of the 5th Workshop on Representation Learning for NLP
Month:
July
Year:
2020
Address:
Online
Editors:
Spandana Gella, Johannes Welbl, Marek Rei, Fabio Petroni, Patrick Lewis, Emma Strubell, Minjoon Seo, Hannaneh Hajishirzi
Venue:
RepL4NLP
SIG:
SIGREP
Publisher:
Association for Computational Linguistics
Note:
Pages:
55–60
Language:
URL:
https://aclanthology.org/2020.repl4nlp-1.8
DOI:
10.18653/v1/2020.repl4nlp-1.8
Bibkey:
Cite (ACL):
Lis Pereira, Xiaodong Liu, Fei Cheng, Masayuki Asahara, and Ichiro Kobayashi. 2020. Adversarial Training for Commonsense Inference. In Proceedings of the 5th Workshop on Representation Learning for NLP, pages 55–60, Online. Association for Computational Linguistics.
Cite (Informal):
Adversarial Training for Commonsense Inference (Pereira et al., RepL4NLP 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.repl4nlp-1.8.pdf
Video:
 http://slideslive.com/38929774
Data
CosmosQAMC-TACO