Generating Textual Adversaries with Minimal Perturbation

• DOI: 10.18653/v1/2022.findings-emnlp.337
• Publication Date: 2023-08-04T20:21:02Z
• Authors: Xingyi Zhao; Lu Zhang; Depeng Xu; Shuhan Yuan
• Keywords: FOS: Computer and information sciences; Computer Science - Computation and Language; 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology; Computation and Language (cs.CL)

Many word-level adversarial attack approaches for textual data have been proposed in recent studies. However, due to the massive search space consisting of combinations candidate words, existing face problem preserving semantics texts when crafting counterparts. In this paper, we develop a novel strategy find with high similarity original while introducing minimal perturbation. The rationale is that expect small perturbation can better preserve semantic meaning texts. Experiments show that, compared state-of-the-art approaches, our approach achieves higher success rates and lower four benchmark datasets.