[1]任建华,赵若涵.基于特征融合的糖尿病命名实体识别[J].中国医学物理学杂志,2024,41(7):890-896.[doi:DOI:10.3969/j.issn.1005-202X.2024.07.016]
　REN Jianhua,ZHAO Ruohan.Diabetes named entity recognition based on feature fusion[J].Chinese Journal of Medical Physics,2024,41(7):890-896.[doi:DOI:10.3969/j.issn.1005-202X.2024.07.016]
点击复制

基于特征融合的糖尿病命名实体识别()
分享到：

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
41卷
期数:
2024年第7期
页码:
890-896
栏目:
医学生物物理
出版日期:
2024-07-25

文章信息/Info

Title:
Diabetes named entity recognition based on feature fusion
文章编号:
1005-202X（2024）07-0890-07
作者:
任建华; 赵若涵
辽宁工程技术大学电子与信息工程学院， 辽宁 葫芦岛 125105
Author(s):
REN Jianhua;  ZHAO Ruohan
School of Electronic and Information Engineering, Liaoning Technical University, Huludao 125105, China
关键词:
糖尿病; 命名实体识别; 特征融合; 注意力机制
Keywords:
Keywords: diabetes named entity recognition feature fusion attention mechanism
分类号:
R318；R587.1
DOI:
DOI:10.3969/j.issn.1005-202X.2024.07.016
文献标志码:
A
摘要:
针对医学糖尿病领域命名实体识别中存在实体种类多样性、数据稀缺等问题，提出了基于特征融合的糖尿病命名实体识别方法。以BERT+BILSTM+CRF为基准模型，在3方面进行改进。首先，使用预训练模型RoBERTa-wwm-ext作为模型嵌入层，提供字符级嵌入，利用其在训练阶段进行全词掩码来获取含有先验知识的语义表示。其次，使用双向长短时记忆网络和迭代膨胀卷积神经网络并行提取特征，以获取不同粒度的特征。同时，结合注意力机制进行动态特征融合，从而更好地理解数据的关键信息，以获得更丰富的上下文特征。最后，采用条件随机场进行解码，获得最终的预测结果。该模型在包含18种实体类别的中文糖尿病数据集DiaKG上的F1值达到了79.58%，实验结果表明，与High-Order MKGragh模型相比，该模型的F1值提升了5.38%，充分说明了特征融合的方法能够有效识别糖尿病实体。
Abstract:
Abstract: To overcome the challenges of entity diversity and data scarcity in diabetes named entity recognition, feature fusion based named entity recognition is proposed. With BERT+BILSTM+CRF as the benchmark model, improvements are made in 3 aspects. (1) The pre-trained model RoBERTa-wwm-ext is introduced as the model embedding layer to provide character-level embedding, and the whole word mask is used in the training stage to obtain semantic representation containing prior knowledge. (2) bidirectional long short-term memory network and iterated dilated convolutional neural network are used to extract features in parallel to obtain features of different granularities. At the same time, the dynamic feature fusion is combined with the attention mechanism to better understand the key information of the data， thus obtaining richer contextual features. (3) The conditional random field is decoded to obtain the final prediction results. The proposed model achieves an F1 value of 79.58% which is 5.38% higher than High-Order MKGraph on DiaKG, a Chinese diabetes data set containing 18 entity categories, fully demonstrating that the feature fusion based method can effectively identify diabetic entities.

相似文献/References:

[1]雍军光,阮 萍,沈洪涛,等.新型显微成像与分析系统在糖尿病患者红细胞定量研究中的应用[J].中国医学物理学杂志,2014,31(04):5081.[doi:10.3969/j.issn.1005-202X.2014.04.023]
[2]赵承奇,雷涛,罗二平,等.脉冲电磁场对糖尿病大鼠外周神经病变的影响[J].中国医学物理学杂志,2013,30(05):4431.[doi:10.3969/j.issn.1005-202X.2013.05.021]
[3]苑旺,梁媛媛,崔黎丽,等.正极性聚丙烯驻极体对糖尿病大鼠皮肤结构的影响[J].中国医学物理学杂志,2015,32(06):835.[doi:doi:10.3969/j.issn.1005-202X.2015.06.016]
　[J].Chinese Journal of Medical Physics,2015,32(7):835.[doi:doi:10.3969/j.issn.1005-202X.2015.06.016]
[4]余丽玲,陈婷,金浩宇,等.基于支持向量机和自回归积分滑动平均模型组合的血糖值预测[J].中国医学物理学杂志,2016,33(4):381.[doi:10.3969/j.issn.1005-202X.2016.04.012]
　[J].Chinese Journal of Medical Physics,2016,33(7):381.[doi:10.3969/j.issn.1005-202X.2016.04.012]
[5]花琦琦,刘新凤,焦青,等.基于SQL Server的糖尿病信息管理与分析系统[J].中国医学物理学杂志,2016,33(11):1183.[doi:10.3969/j.issn.1005-202X.2016.11.020]
　[J].Chinese Journal of Medical Physics,2016,33(7):1183.[doi:10.3969/j.issn.1005-202X.2016.11.020]
[6]楚轶,冯品,张薇,等. 稳恒磁场刺激对糖尿病动脉粥样硬化大鼠血清和主动脉中VEGF、TGF-β1、TNF-α和IL-6表达的影响[J].中国医学物理学杂志,2017,34(10):1045.[doi:DOI:10.3969/j.issn.1005-202X.2017.10.016]
　[J].Chinese Journal of Medical Physics,2017,34(7):1045.[doi:DOI:10.3969/j.issn.1005-202X.2017.10.016]
[7]郭一冰,崔栋,薛雅卓,等. 糖尿病行为量表计算工具箱的研制[J].中国医学物理学杂志,2018,35(2):195.[doi:DOI:10.3969/j.issn.1005-202X.2018.02.015]
　GUO Yibing,CUI Dong,XUE Yazhuo,et al. Development of diabetes behavioral scale calculation toolbox[J].Chinese Journal of Medical Physics,2018,35(7):195.[doi:DOI:10.3969/j.issn.1005-202X.2018.02.015]
[8]张嘉阳,黄河,刘子怡,等. 基于Gabor滤波器的糖尿病视网膜新生血管检测[J].中国医学物理学杂志,2018,35(8):968.[doi:DOI:10.3969/j.issn.1005-202X.2018.08.019]
　ZHANG Jiayang,HUANG He,LIU Ziyi,et al. Gabor filter-based detection of neovascularization due to diabetic retinopathy[J].Chinese Journal of Medical Physics,2018,35(7):968.[doi:DOI:10.3969/j.issn.1005-202X.2018.08.019]
[9]陈真诚,杜莹,邹春林,等. 基于K-Nearest Neighbor和神经网络的糖尿病分类研究[J].中国医学物理学杂志,2018,35(10):1220.[doi:DOI:10.3969/j.issn.1005-202X.2018.010.022]
　CHEN Zhencheng,DU Ying,ZOU Chunlin,et al. Classification of diabetes based on K-Nearest Neighbor and neural network[J].Chinese Journal of Medical Physics,2018,35(7):1220.[doi:DOI:10.3969/j.issn.1005-202X.2018.010.022]
[10]张迪,陈真诚,梁永波,等. 协同训练算法在无创血糖检测中的应用[J].中国医学物理学杂志,2018,35(11):1295.[doi:DOI:10.3969/j.issn.1005-202X.2018.11.011]
　ZHANG Di,CHEN Zhencheng,LIANG Yongbo,et al. Application of co-training algorithm in noninvasive blood glucose detection[J].Chinese Journal of Medical Physics,2018,35(7):1295.[doi:DOI:10.3969/j.issn.1005-202X.2018.11.011]
[11]马诗语,黄润才.基于ALBERT与BILSTM的糖尿病命名实体识别[J].中国医学物理学杂志,2021,38(11):1438.[doi:DOI:10.3969/j.issn.1005-202X.2021.11.021]
　MA Shiyu,HUANG Runcai.Named entity recognition of diabetes based on ALBERT and BILSTM[J].Chinese Journal of Medical Physics,2021,38(7):1438.[doi:DOI:10.3969/j.issn.1005-202X.2021.11.021]

备注/Memo

备注/Memo:
【收稿日期】2024-02-20 【基金项目】国家自然科学基金（61772249） 【作者简介】任建华，硕士，副教授，研究方向：智能数据处理、数据库理论及应用，E-mail: renjh4665@163.com；赵若涵，硕士研究生，主要研究方向：自然语言处理，E-mail: 2689613378@qq.com
更新日期/Last Update: 2024-07-13