| Summary: | Chinese nested named entity recognition(CNNER) is a challenging task due to the absence of natural delimiters in Chinese and the complexity of the nested structure.In this paper,we propose a novel boundary-aware layered neural model(BLNM) with segmentation attention for the CNNER task.To exploit some semantic relation among adjacent characters,we first design a segmentation attention network to capture the potential word information and enhance character representation.Next,we model the nested structure with dynamically stacked Flat NER networks to detect entities in an inner to outer manner.We also design a boundary generative module to connect adjacent Flat NER layers,which can mark the boundary and position of detected entities and greatly alleviate the error propagation problem.Experiment results on ACE 2005 Chinese nested NE dataset show that the proposed model achieves superior performance than the state-of-the-art methods.
|
|---|