引入词集级注意力机制的中文命名实体识别方法

Chinese word segmentation is necessary to provide word-level information for Chinese named entity recognition. Recently, many researchers have attempted to improve the performance of Chinese named entity recognition by incorporating word-level information into character representations. However, they ignored the different importance among word sets. Aiming at this issue, this paper focuses on word-set level relationship and proposes a new method, which incorporates word-set attention into Chinese named entity recognition, that adaptively recalibrates word-set level features by attention mechanism. This method makes the network perform better in Chinese named entity recognition by learning to selectively emphasize informative features and suppress useless ones. Finally, the effectiveness of the method is verified by experiments on multiple datasets.