CEPTNER: Contrastive learning Enhanced Prototypical network for Two-stage few-shot Named Entity Recognition

Enze Zha, De-Long Zeng, Man Lin, Ying Shen

Abstract: Highlights•A two-stage approach named CEPTNER is proposed to address few-shot NER problem.•Designing a span filter helps to remove false generated spans.•Utilizing contrastive learning methods for hidden relationship exploring is effective.