SSMBA: Self-Supervised Manifold Based Data Augmentation for Improving Out-of-Domain Robustness
This addresses the issue of poor out-of-domain robustness for natural language processing models, though it is incremental as it builds on existing data augmentation and manifold assumptions.
The paper tackled the problem of models failing to generalize to out-of-domain data by introducing SSMBA, a data augmentation method that generates synthetic examples using corruption and reconstruction functions to stay on the data manifold, resulting in gains such as 0.8% accuracy on OOD Amazon reviews and 1.8% accuracy on OOD MNLI.
Models that perform well on a training domain often fail to generalize to out-of-domain (OOD) examples. Data augmentation is a common method used to prevent overfitting and improve OOD generalization. However, in natural language, it is difficult to generate new examples that stay on the underlying data manifold. We introduce SSMBA, a data augmentation method for generating synthetic training examples by using a pair of corruption and reconstruction functions to move randomly on a data manifold. We investigate the use of SSMBA in the natural language domain, leveraging the manifold assumption to reconstruct corrupted text with masked language models. In experiments on robustness benchmarks across 3 tasks and 9 datasets, SSMBA consistently outperforms existing data augmentation methods and baseline models on both in-domain and OOD data, achieving gains of 0.8% accuracy on OOD Amazon reviews, 1.8% accuracy on OOD MNLI, and 1.4 BLEU on in-domain IWSLT14 German-English.