CrossNER: Evaluating Cross-Domain Named Entity Recognition

Cross-domain named entity recognition (NER) models are able to cope with the scarcity issue of NER samples in target domains. However, most existing benchmarks lack domain-specialized types or do not focus on a certain domain, leading less effective cross-domain evaluation. To address these obstacles, we introduce dataset (CrossNER), fully-labeled collection data spanning over five diverse domains specialized categories for different Additionally, also provide domain-related corpus since using it continue pre-training language (domain-adaptive pre-training) is domain adaptation. We then conduct comprehensive experiments explore effectiveness leveraging levels and strategies domain-adaptive task. Results show that focusing fractional containing entities utilizing more challenging strategy beneficial adaptation, our proposed method can consistently outperform baselines. Nevertheless, illustrate challenge this hope baselines will catalyze research adaptation area. The code available at https://github.com/zliucr/CrossNER.