Context-aware Adversarial Training for Name Regularity Bias in Named Entity Recognition

Abstract In this work, we examine the ability of NER models to use contextual information when predicting type an ambiguous entity. We introduce NRB, a new testbed carefully designed diagnose Name Regularity Bias models. Our results indicate that all state-of-the-art tested show such bias; BERT fine-tuned significantly outperforming feature-based (LSTM-CRF) ones on despite having comparable (sometimes lower) performance standard benchmarks. To mitigate bias, propose novel model-agnostic training method adds learnable adversarial noise some entity mentions, thus enforcing focus more strongly signal, leading significant gains NRB. Combining it with two other strategies, data augmentation and parameter freezing, leads further gains.