Abstract
In this paper, we explore the following question: Are synonym substitution attacks really synonym substitution attacks (SSAs)?We approach this question by examining how SSAs replace words in the original sentence and show that there are still unresolved obstacles that make current SSAs generate invalid adversarial samples. We reveal that four widely used word substitution methods generate a large fraction of invalid substitution words that are ungrammatical or do not preserve the original sentence’s semantics. Next, we show that the semantic and grammatical constraints used in SSAs for detecting invalid word replacements are highly insufficient in detecting invalid adversarial samples.- Anthology ID:
- 2023.findings-acl.117
- Volume:
- Findings of the Association for Computational Linguistics: ACL 2023
- Month:
- July
- Year:
- 2023
- Address:
- Toronto, Canada
- Editors:
- Anna Rogers, Jordan Boyd-Graber, Naoaki Okazaki
- Venue:
- Findings
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 1853–1878
- Language:
- URL:
- https://aclanthology.org/2023.findings-acl.117
- DOI:
- 10.18653/v1/2023.findings-acl.117
- Cite (ACL):
- Cheng-Han Chiang and Hung-yi Lee. 2023. Are Synonym Substitution Attacks Really Synonym Substitution Attacks?. In Findings of the Association for Computational Linguistics: ACL 2023, pages 1853–1878, Toronto, Canada. Association for Computational Linguistics.
- Cite (Informal):
- Are Synonym Substitution Attacks Really Synonym Substitution Attacks? (Chiang & Lee, Findings 2023)
- PDF:
- https://preview.aclanthology.org/dois-2013-emnlp/2023.findings-acl.117.pdf