Table 10 Performance of XLM-RSA under adversarial perturbations.
From: Multilingual sentiment analysis in restaurant reviews using aspect focused learning
Perturbation type | Accuracy (%) | Precision (%) | Recall (%) | F1-score (%) |
|---|---|---|---|---|
No perturbation (baseline) | 92.3 | 91.5 | 92.0 | 91.7 |
Character-level noise | 90.8 | 90.1 | 90.5 | 90.3 |
Word substitution | 91.2 | 90.7 | 91.0 | 90.8 |
Word deletion | 89.6 | 89.0 | 89.3 | 89.1 |
