正在加载图片...
474 工程科学学报,第42卷,第4期 表5DLAM与现有模型结果对比 [7 Zhang X W,LiZ.Chinese electronic medical record named entity recognition based on multi-feature fusion.Softw Guide,2017, Table 5 Comparison of DLAM and existing model results o 16(2):128 Model Marco-P Marco-R Marco-F1 (张祥伟,李智.基于多特征融合的中文电子病历命名实体识别. CRF multi-features7 92.03 87.09 89.49 软件导刊,2017,16(2):128) BiLSTM-CRFR网 91.12 89.74 90.43 [8] Yu L,Jin L Z,Wang M F,et al.Recognition of human hypoxic DLAM 96.70 97.70 97.20 state based on deep learning.Chin J Eng,2019,41(6):817 (于露,金龙哲,王梦飞,等.基于深度学习的人体低氧状态识别 清的语言特点,对中文电子病历中的四类实体- 工程科学学报,2019,41(6):817) 疾病、症状、药品、操作进行命名实体识别研究 [9] Xia Y B,Zhen J L,Zhao Y F,et al.Deep learning based named (1)结合电子病历文本通过统计分析构建了 entity recognition of electronic medical record.Electron Sci 个小规模的医疗领域词典. Technol,2018,31(11:31 (夏宇彬,郑建立,赵逸凡,等基于深度学习的电子病历命名实 (2)将经典序列标注算法CF与富含领域知 体识别.电子科技,2018.31(11):31) 识的词典相结合,提出了一种预标注-二次标注的 [10]Li F,Zhang M S,Tian B,et al.Recognizing irregular entities in 双层标注模型DLAM.通过一次预标注-二次精确 biomedical text via deep neural networks.Pattern Recognit Lett, 标注两种不同粒度的标注完成对中文医疗实体的 2018,105:105 识别.经过实验验证,DLAM在测试集上的宏精确 [11]Liu Z J,Yang M,Wang X L,et al.Entity recognition from clinical 率为96.7%、宏召回率为97.7%、宏F1值为97.2%, texts via recurrent neural networks.BMC Med Inf Decis Making, 可准确地对中文医疗实体进行识别 2017,17(Suppl2):67 (3)对比分析采用注意力机制的深度神经网络 [12]Chowdhury S,Dong X S,Qian L J,et al.A multitask bi- directional RNN model for named entity recognition on Chinese 的识别效果,结果表明提出的双层标注模型DLAM electronic medical records.BMC Bioinf,2018,19(Suppl 17):499 在测试数据集上表现优越于深度神经网络 [13]Shen Z.Named Entity Recognition for Chinese Electronic Record with Neural Nenwork[Dissertation].Beijing:Beijing University of 参考文献 Posts and Telecommunications,2018 1]Zhang L B.Word Segmentation and Named Entity Mining Based (申站基于神经网络的中文电子病历命名实体识别[学位论文] on Semi Supervised Learning for Chinese EMR[Dissertation] 北京:北京邮电大学,2018) Harbin:Harbin Institute of Technology,2014 [14]Wei QK,Chen T,Xu R F,et al.Disease named entity recognition (张立邦.基于半监督学习的中文电子病历分词和名实体挖掘 by combining conditional random fields and bidirectional recurrent [学位论文].哈尔滨:哈尔滨工业大学,2014) neural networks.Database,2016,2016:baw140 [2]Huang Z H,Xu W,Yu K.Bidirectional LSTM-CRF Models for [15]Wu Y H,Yang X,Bian J,et al.Combine factual medical Sequence Tagging[J/OL].arXiv preprint.(2015-08-09)[2019-09- knowledge and distributed word representation to improve clinical 04].https://arxiv.org/abs/1508.01991 named entity recognition.AML4 Annu Symp Proc,2018,2018: [3]Wang Y Q,Yu Z H,Chen L et al.Supervised methods for 1110 symptom name recognition in free-text clinical records of [16]Jagannatha A N,Yu H.Bidirectional RNN for medical event traditional Chinese medicine:an empirical study.J Biomed Inf, detection in electronic health records /Proceedings of the 2016 2014,47:91 Conference of the North American Chapter of the Association for [4]Xu Y,Wang Y N,Liu T R,et al.Joint segmentation and named Computational Linguistics:Human Language Technologies. entity recognition using dual decomposition in Chinese discharge California,2016:473 summaries.J Am Med Inf Assoc,2014,21(el):e84 [17]Rajkomar A,Oren E,Chen K,et al.Scalable and accurate deep [5]Lei J B,Tang BZ,Lu X Q,et al.A comprehensive study of named learning with electronic health records[J/OL].arXiv preprint. entity recognition in Chinese clinical text.J Am Med Inf Assoc, (2018-05-11)[2019-09-041.htps://arxiv.org/abs/1801.07860 2014,21(5):808 [18]Wang Y,Wang L,Rastegar-Mojarad M,et al.Clinical information [6] Xu Y,Ge YQ.Wang Q,et al.Medical name entity recognition extraction applications:a literature review.JBiomed Inf,2018,77: and application in Chinese admission record of stroke patients 34 based on CRF and RUTA rule.J Sun Yat-sen Univ Med Sci,2018 [19]Luka G,Andrey K,Paul G,et al.Named entity recognition in 39(3):455 electronic health records using transfer learning bootstrapped (许源,葛艳秋,王强,等.基于CRF与RUTA规则相结合的卒中 neural networks[J/OL].arYry preprint.(2019-07-29)[2019-09- 入院记录医学实体识别及应用.中山大学学报(医学版),2018 04].https://arxiv.org/abs/1901.01592 39(3):455) [20]Li W,Zhao D Z,Li B,et al.Combining CRF and rule based清的语言特点,对中文电子病历中的四类实体—— 疾病、症状、药品、操作进行命名实体识别研究. (1)结合电子病历文本通过统计分析构建了 一个小规模的医疗领域词典. (2)将经典序列标注算法 CRF 与富含领域知 识的词典相结合,提出了一种预标注–二次标注的 双层标注模型 DLAM. 通过一次预标注-二次精确 标注两种不同粒度的标注完成对中文医疗实体的 识别. 经过实验验证,DLAM 在测试集上的宏精确 率为 96.7%、宏召回率为 97.7%、宏 F1 值为 97.2%, 可准确地对中文医疗实体进行识别. (3)对比分析采用注意力机制的深度神经网络 的识别效果,结果表明提出的双层标注模型 DLAM 在测试数据集上表现优越于深度神经网络. 参    考    文    献 Zhang L B. Word Segmentation and Named Entity Mining Based on Semi Supervised Learning for Chinese EMR[Dissertation]. Harbin: Harbin Institute of Technology, 2014 (张立邦. 基于半监督学习的中文电子病历分词和名实体挖掘 [学位论文]. 哈尔滨: 哈尔滨工业大学, 2014) [1] Huang  Z  H,  Xu  W,  Yu  K.  Bidirectional  LSTM-CRF  Models  for Sequence Tagging[J/OL]. arXiv preprint. (2015-08-09) [2019-09- 04]. https://arxiv.org/abs/1508.01991 [2] Wang  Y  Q,  Yu  Z  H,  Chen  L,  et  al.  Supervised  methods  for symptom  name  recognition  in  free-text  clinical  records  of traditional  Chinese  medicine:  an  empirical  study. J Biomed Inf, 2014, 47: 91 [3] Xu Y, Wang Y N, Liu T R, et al. Joint segmentation and named entity recognition using dual decomposition in Chinese discharge summaries. J Am Med Inf Assoc, 2014, 21(e1): e84 [4] Lei J B, Tang B Z, Lu X Q, et al. A comprehensive study of named entity  recognition  in  Chinese  clinical  text. J Am Med Inf Assoc, 2014, 21(5): 808 [5] Xu  Y,  Ge  Y  Q,  Wang  Q,  et  al.  Medical  name  entity  recognition and  application  in  Chinese  admission  record  of  stroke  patients based on CRF and RUTA rule. J Sun Yat-sen Univ Med Sci, 2018, 39(3): 455 (许源, 葛艳秋, 王强, 等. 基于CRF与RUTA规则相结合的卒中 入院记录医学实体识别及应用. 中山大学学报(医学版), 2018, 39(3):455) [6] Zhang X W, Li Z. Chinese electronic medical record named entity recognition  based  on  multi-feature  fusion. Softw Guide,  2017, 16(2): 128 (张祥伟, 李智. 基于多特征融合的中文电子病历命名实体识别. 软件导刊, 2017, 16(2):128) [7] Yu L, Jin L Z, Wang M F, et al. Recognition of human hypoxic state based on deep learning. Chin J Eng, 2019, 41(6): 817 (于露, 金龙哲, 王梦飞, 等. 基于深度学习的人体低氧状态识别. 工程科学学报, 2019, 41(6):817) [8] Xia Y B, Zhen J L, Zhao Y F, et al. Deep learning based named entity  recognition  of  electronic  medical  record. Electron Sci Technol, 2018, 31(11): 31 (夏宇彬, 郑建立, 赵逸凡, 等. 基于深度学习的电子病历命名实 体识别. 电子科技, 2018, 31(11):31) [9] Li F, Zhang M S, Tian B, et al. Recognizing irregular entities in biomedical  text  via  deep  neural  networks. Pattern Recognit Lett, 2018, 105: 105 [10] Liu Z J, Yang M, Wang X L, et al. Entity recognition from clinical texts via recurrent neural networks. BMC Med Inf Decis Making, 2017, 17(Suppl 2): 67 [11] Chowdhury  S,  Dong  X  S,  Qian  L  J,  et  al.  A  multitask  bi￾directional  RNN  model  for  named  entity  recognition  on  Chinese electronic medical records. BMC Bioinf, 2018, 19(Suppl 17): 499 [12] Shen Z. Named Entity Recognition for Chinese Electronic Record with Neural Network[Dissertation]. Beijing: Beijing University of Posts and Telecommunications, 2018 (申站.基于神经网络的中文电子病历命名实体识别[学位论文]. 北京: 北京邮电大学, 2018) [13] Wei Q K, Chen T, Xu R F, et al. Disease named entity recognition by combining conditional random fields and bidirectional recurrent neural networks. Database, 2016, 2016: baw140 [14] Wu  Y  H,  Yang  X,  Bian  J,  et  al.  Combine  factual  medical knowledge and distributed word representation to improve clinical named  entity  recognition. AMIA Annu Symp Proc,  2018,  2018: 1110 [15] Jagannatha  A  N,  Yu  H.  Bidirectional  RNN  for  medical  event detection  in  electronic  health  records  // Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. California, 2016: 473 [16] Rajkomar  A,  Oren  E,  Chen  K,  et  al.  Scalable  and  accurate  deep learning  with  electronic  health  records[J/OL]. arXiv preprint. (2018-05-11) [2019-09-04]. https://arxiv.org/abs/1801.07860 [17] Wang Y, Wang L, Rastegar-Mojarad M, et al. Clinical information extraction applications: a literature review. J Biomed Inf, 2018, 77: 34 [18] Luka  G,  Andrey  K,  Paul  G,  et  al.  Named  entity  recognition  in electronic  health  records  using  transfer  learning  bootstrapped neural  networks[J/OL]. arXiv preprint.  (2019-07-29)  [2019-09- 04]. https://arxiv.org/abs/1901.01592 [19] [20] Li  W,  Zhao  D  Z,  Li  B,  et  al.  Combining  CRF  and  rule  based 表 5    DLAM 与现有模型结果对比 Table 5    Comparison of DLAM and existing model results % Model Marco-P Marco-R Marco-F1 CRF_multi-features[27] 92.03 87.09 89.49 BiLSTM-CRF[27] 91.12 89.74 90.43 DLAM 96.70 97.70 97.20 · 474 · 工程科学学报,第 42 卷,第 4 期
<<向上翻页向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有