On the Robustness of Offensive Language Classifiers

Jonathan Rusert, Zubair Shafiq, Padmini Srinivasan


Abstract
Social media platforms are deploying machine learning based offensive language classification systems to combat hateful, racist, and other forms of offensive speech at scale. However, despite their real-world deployment, we do not yet comprehensively understand the extent to which offensive language classifiers are robust against adversarial attacks. Prior work in this space is limited to studying robustness of offensive language classifiers against primitive attacks such as misspellings and extraneous spaces. To address this gap, we systematically analyze the robustness of state-of-the-art offensive language classifiers against more crafty adversarial attacks that leverage greedy- and attention-based word selection and context-aware embeddings for word replacement. Our results on multiple datasets show that these crafty adversarial attacks can degrade the accuracy of offensive language classifiers by more than 50% while also being able to preserve the readability and meaning of the modified text.
Anthology ID:
2022.acl-long.513
Volume:
Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
May
Year:
2022
Address:
Dublin, Ireland
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
7424–7438
Language:
URL:
https://aclanthology.org/2022.acl-long.513
DOI:
10.18653/v1/2022.acl-long.513
Bibkey:
Cite (ACL):
Jonathan Rusert, Zubair Shafiq, and Padmini Srinivasan. 2022. On the Robustness of Offensive Language Classifiers. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7424–7438, Dublin, Ireland. Association for Computational Linguistics.
Cite (Informal):
On the Robustness of Offensive Language Classifiers (Rusert et al., ACL 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.acl-long.513.pdf
Software:
 2022.acl-long.513.software.zip
Code
 jonrusert/robustnessofoffensiveclassifiers
Data
OLID