摘要
随着信息技术的发展,电子文档在糖尿病领域的信息记录中得到了大量应用,通过自动化技术对这些电子文档进行分析具有重大的意义。由于现有的命名实体识别方法在糖尿病领域中识别准确率偏低。为了改变这种现状,提出了双层的双向长短时记忆神经网络条件随机场模型(Bi-LSTM-CRF),并将其应用到糖尿病领域命名实体识别任务中。实验结果表明该模型在包含15种实体类别的数据集上准确率达到了89.14%,且在外部测试集上平均F 1值为72.89%,充分揭示了双层Bi-LSTM-CRF模型的有效性。
With the development of information technology,electronic documents have been widely used in the information record of diabetes.Analysis of these electronic documents through automation technology has a great significance.Due to the low accuracy of existing named entity recognition methods in the field of diabetes,a double-layer bidirectional long-short-term memory neural network conditional random field model(Bi-LSTM-CRF)was proposed and applied to the task of named entity recognition in the field of diabetes.Experimental results show that the accuracy of the model is 89.14%on a dataset containing 15 entity categories,and the average F 1 score on the external test dataset is 72.89%,which fully reveals the effectiveness of the double-layer Bi-LSTM-CRF model.
作者
何春辉
王梦贤
何小波
HE Chunhui;WANG Mengxian;HE Xiaobo(School of Mathematics and Computational Sciences,Xiangtan University,Xiangtan 411105,China;College of Management,Hunan City University,Yiyang 413000,China;Troop of 75841,Changsha 410000,China)
出处
《邵阳学院学报(自然科学版)》
2020年第1期21-26,共6页
Journal of Shaoyang University:Natural Science Edition
基金
湖南省教育厅科研项目(17C0293)