RoCBert: Robust Chinese Bert with Multimodal Contrastive Pretraining
Published in ACL, 2022
Large-scale pretrained language models have achieved SOTA results on NLP tasks. However, they have been shown vulnerable to adversarial attacks especially for logographic languages like Chinese. In this work, we propose ROCBERT: a pretrained Chinese Bert that is robust to various forms of adversarial attacks like word perturbation, synonyms, typos, etc. It is pretrained with the contrastive learning objective which maximizes the label consistency under different synthesized adversarial examples. The model takes as input multimodal information including the semantic, phonetic and visual features. We show all these features are important to the model robustness since the attack can be performed in all the three forms. Across 5 Chinese NLU tasks, ROCBERT outperforms strong baselines under three blackbox adversarial algorithms without sacrificing the performance on clean testset. It also performs the best in the toxic content detection task under human-made attacks.