MCL: Multi-Granularity Contrastive Learning Framework for Chinese NER

Recently, researchers have applied the word-character lattice framework to integrated word information, which has become very popular for Chinese named entity recognition (NER). However, prior approaches fuse information by different variants of encoders such as Lattice LSTM or Flat-Lattice Transformer, but are still not data-efficient indeed fully grasp depth interaction cross-granularity and important from lexicon. In this paper, we go beyond typical structure propose a novel Multi-Granularity Contrastive Learning (MCL), that aims optimize inter-granularity distribution distance emphasize critical matched words in By carefully combining contrastive learning bi-granularity learning, network can explicitly leverage lexicon on initial structure, further provide more dense interactions across-granularity, thus significantly improving model performance. Experiments four NER datasets show MCL obtains state-of-the-art results while considering efficiency. The source code proposed method is publicly available at https://github.com/zs50910/MCL