摘要
作为众多任务的子任务,命名实体识别的发展较为迅速,但在中文命名实体识别领域,还存在不少问题,嵌套实体就是其中一个难点。文中根据结点的传入和传出,使用了图卷积神经网络提取图特征,改善了嵌套实体的准确度,并通过使用图神经网络处理中文命名实体识别的问题,更好地融合了词典信息。另外,文中分别对两类数据集进行了实验验证,结果显示,相比其他模型,该模型有所提高。
As a subtask of many tasks,named entity recognition has developed rapidly,but there are still many problems in the field of Chinese named entity recognition,and nested entities are one of the difficulties.This paper uses graph convolutional neural networks to extract graph features according to the incoming and outgoing nodes,improves the accuracy of nested entities,and by using graph neural networks to deal with the problem of Chinese named entity recognition,so as to better integrate dictionary information.In addition,experimental verification of two types of datasets is carried out in this paper,and the results show that the model has improved compared with other models.
作者
曲晓东
李佳昊
QU Xiaodong;LI Jiahao(Zhongyuan University of Technology,Zhengzhou 450007,China)
出处
《移动信息》
2023年第6期234-236,共3页
MOBILE INFORMATION
关键词
嵌套问题
中文命名
实体识别
Nesting problem
Chinese name
Entity identification