A multi-granularity in-context learning method for few-shot Named Entity Recognition via Knowledgeable Parameters Fine-tuning

Qihui Zhao, Tianhan Gao, Nan Guo

Abstract: Highlights•We propose a novel multi-granularity in-context learning-based method for few-shot NER.•We design a parameters fine-tuning method to store and utilize NER knowledge.•We use contrastive-learning to better distinguish entity boundaries from other words.•Our method excels in few-shot NER while matching SOTA in rich-resource settings.