针对目前自动ICD(international classification of diseases)编码任务存在标签空间大、诊断代码分布不均衡与临床文本表征差的问题,提出一种融合Longformer与标签注意力的分层ICD自动编码模型。借助Clinical-Longformer预训练语言模型...针对目前自动ICD(international classification of diseases)编码任务存在标签空间大、诊断代码分布不均衡与临床文本表征差的问题,提出一种融合Longformer与标签注意力的分层ICD自动编码模型。借助Clinical-Longformer预训练语言模型获得融合长文本语境的词向量表征。通过将标签的语义表示与注意力机制相结合,捕捉临床文本中与诊断代码相关的关键特征信息,获取更精准的文本表示。引入分层联合学习机制,建立分层预测层解码输出ICD编码。实验结果表明,该模型的准确率、召回率与F1值均高于现有模型,验证了该方法进行自动ICD编码的有效性,为实施疾病诊断相关分组提供高质量的数据支撑。展开更多
A multi-level evaluation model for the superstructure of a damaged prestressed concrete girder or beam bridge is established, and the evaluation indices of the model as well as the rating standards are defined. A norm...A multi-level evaluation model for the superstructure of a damaged prestressed concrete girder or beam bridge is established, and the evaluation indices of the model as well as the rating standards are defined. A normal relative function about the evaluation indices of each element is developed to calculate the relative degree, and for each element there are no sub-level elements. When evaluating the elements in the sub-item level or the index level of the model, the weights of elements pertain to one adopted element, taking into account their degrees of deterioration. Since the relative degrees and structure evaluation scales on the damage conditions are applied to characterize the superstructure of damaged prestressed concrete girder bridges, this method can evaluate the prestressed structure in detail, and the evaluation results agree with the Code for Maintenance of Highway Bridges and Culvers （JTG Hll--2004 ）. Finally, a bridge in Jilin province is taken as an example, using the method developed to evaluate its damage conditions, which gives an effective way for bridge engineering.展开更多
文摘针对目前自动ICD(international classification of diseases)编码任务存在标签空间大、诊断代码分布不均衡与临床文本表征差的问题,提出一种融合Longformer与标签注意力的分层ICD自动编码模型。借助Clinical-Longformer预训练语言模型获得融合长文本语境的词向量表征。通过将标签的语义表示与注意力机制相结合,捕捉临床文本中与诊断代码相关的关键特征信息,获取更精准的文本表示。引入分层联合学习机制,建立分层预测层解码输出ICD编码。实验结果表明,该模型的准确率、召回率与F1值均高于现有模型,验证了该方法进行自动ICD编码的有效性,为实施疾病诊断相关分组提供高质量的数据支撑。
文摘A multi-level evaluation model for the superstructure of a damaged prestressed concrete girder or beam bridge is established, and the evaluation indices of the model as well as the rating standards are defined. A normal relative function about the evaluation indices of each element is developed to calculate the relative degree, and for each element there are no sub-level elements. When evaluating the elements in the sub-item level or the index level of the model, the weights of elements pertain to one adopted element, taking into account their degrees of deterioration. Since the relative degrees and structure evaluation scales on the damage conditions are applied to characterize the superstructure of damaged prestressed concrete girder bridges, this method can evaluate the prestressed structure in detail, and the evaluation results agree with the Code for Maintenance of Highway Bridges and Culvers （JTG Hll--2004 ）. Finally, a bridge in Jilin province is taken as an example, using the method developed to evaluate its damage conditions, which gives an effective way for bridge engineering.
