为提升民航应急管理能力,研究基于《中国民用航空应急管理规定》问卷数据,运用轻量级双向编码器表征模型(A Lite Bidirectional Encoder Representations from Transformers,ALBERT)和潜在狄利克雷分配模型(Latent Dirichlet Allocation...为提升民航应急管理能力,研究基于《中国民用航空应急管理规定》问卷数据,运用轻量级双向编码器表征模型(A Lite Bidirectional Encoder Representations from Transformers,ALBERT)和潜在狄利克雷分配模型(Latent Dirichlet Allocation,LDA)进行文档类别划分,结合K-means和均匀流形逼近与投影方法(Uniform Manifold Approximation and Projection,UMAP)进行文档聚类及可视化,根据复杂网络理论构建关键词共现网络,分析网络特性和节点联系,应用逼近理想解排序法(Technique for Order Preference by Similarity to an Ideal Solution,TOPSIS)获得节点重要性排序。研究结果显示:网络的平均聚类系数为0.71~0.965,覆盖了不同部门、岗位和环境,代表了民航应急管理的普遍情况;在人员兼职、空管单位、支线机场等应急体系建设存在不足的情况下,建议民航系统通过增设专职应急管理机构、优化资源配置、加强培训支持等方式提升应急管理能力。研究结果可为民航应急体系建设提供数据支持。展开更多
中文电子病历命名实体识别主要是研究电子病历病程记录文书数据集,文章提出对医疗手术麻醉文书数据集进行命名实体识别的研究。利用轻量级来自Transformer的双向编码器表示(A Lite Bidirectional Encoder Representation from Transform...中文电子病历命名实体识别主要是研究电子病历病程记录文书数据集,文章提出对医疗手术麻醉文书数据集进行命名实体识别的研究。利用轻量级来自Transformer的双向编码器表示(A Lite Bidirectional Encoder Representation from Transformers,ALBERT)预训练模型微调数据集和Tranfomers中的trainer训练器训练模型的方法,实现在医疗手术麻醉文书上识别手术麻醉事件命名实体与获取复杂麻醉医疗质量控制指标值。文章为医疗手术麻醉文书命名实体识别提供了可借鉴的思路,并且为计算复杂麻醉医疗质量控制指标值提供了一种新的解决方案。展开更多
针对现有的数字化档案多标签分类方法存在分类标签之间缺少关联性的问题,提出一种用于档案多标签分类的深层神经网络模型ALBERT-Seq2Seq-Attention.该模型通过ALBERT(A Little BERT)预训练语言模型内部多层双向的Transfomer结构获取进...针对现有的数字化档案多标签分类方法存在分类标签之间缺少关联性的问题,提出一种用于档案多标签分类的深层神经网络模型ALBERT-Seq2Seq-Attention.该模型通过ALBERT(A Little BERT)预训练语言模型内部多层双向的Transfomer结构获取进行文本特征向量的提取,并获得上下文语义信息;将预训练提取的文本特征作为Seq2Seq-Attention(Sequence to Sequence-Attention)模型的输入序列,构建标签字典以获取多标签间的关联关系.将分类模型在3种数据集上分别进行对比实验,结果表明:模型分类的效果F1值均超过90%.该模型不仅能提高档案文本的多标签分类效果,也能关注标签之间的相关关系.展开更多
Introduction: Seizures are one of the most common neurological complications in the infant period. The aim of our study was to describe the epidemiological, clinical, therapeutic and prognostic features of seizures in...Introduction: Seizures are one of the most common neurological complications in the infant period. The aim of our study was to describe the epidemiological, clinical, therapeutic and prognostic features of seizures in infants at the Albert Royer Children’s Hospital (Senegal). Materials and Methods: This was a retrospective, descriptive study from 1 January 2012 to 30 September 2018 of infants aged 0 days to 2 months who presented with seizures. Results: The hospital rate was 8.5%. Almost all the mothers (99.1%) had undergone at least 3 antenatal visits. Urogenital infection, gestational arterial hypertension and funicular anomalies were the main pregnancy-related pathologies. Delivery was vaginal in the majority of cases (80.9%). Most infants (43.6%) had not cried at birth. The majority of infants (63%) were born at term. Trophicity was normal in 68% of cases. The average age of the infants was 6.7 days. The main causes of seizures were hypoxic-ischemic encephalopathy (48.7%), metabolic disturbances (48.1%) and central ոеrvοսѕ system infections (15.6%). Phenobarbital was the 1st-line anticonvulsant. The case fatality rate was 39.5%. The main sequela observed were delayed psychomotor development (20.6%). Conclusion: Optimal management of infant seizures requires early diagnosis and etiological treatment by improving the quality of perinatal care to ensure better management of risk factors, as well as increasing the availability of neuroimaging equipment.展开更多
园林植物知识图谱可为顾及区域适应性、观赏性和生态性等因子的绿化树种的选型提供知识支持。植物描述文本的实体识别及关系抽取是知识图谱构建的关键环节。针对植物领域未有公开的标注数据集,本文阐述了园林植物数据集的构建流程,定义...园林植物知识图谱可为顾及区域适应性、观赏性和生态性等因子的绿化树种的选型提供知识支持。植物描述文本的实体识别及关系抽取是知识图谱构建的关键环节。针对植物领域未有公开的标注数据集,本文阐述了园林植物数据集的构建流程,定义了园林植物的概念体系结构,完成了园林植物语料库的构建。针对现有Word2vec、ELMo和BERT等语言模型存在无法解决多义词、融合上下文能力差、运行速度慢等缺点,提出了嵌入ALBERT(A Lite BERT)预训练语言模型的实体识别和关系抽取模型。ALBERT预训练的动态词向量能够有效地表示文本特征,将其分别输入到BiGRU-CRF命名实体识别模型和BiGRU-Attention关系抽取模型中进行训练,进一步提升实体识别和关系抽取的效果。在园林植物语料库上进行方法的有效性验证,结果表明ALBERT-BiGRU-CRF命名实体识别模型的F1值为0.9517,ALBERT-BiGRU-Attention关系抽取模型的F1值为0.9161,相较于经典的语言模型(如Word2vec、ELMo和BERT等)性能有较为显著的提升。因此基于ALBERT模型的实体与关系抽取任务能有效提高识别分类效果,可将其应用于植物描述文本的实体关系抽取任务中,为园林植物知识图谱自动构建提供方法。展开更多
文摘为提升民航应急管理能力,研究基于《中国民用航空应急管理规定》问卷数据,运用轻量级双向编码器表征模型(A Lite Bidirectional Encoder Representations from Transformers,ALBERT)和潜在狄利克雷分配模型(Latent Dirichlet Allocation,LDA)进行文档类别划分,结合K-means和均匀流形逼近与投影方法(Uniform Manifold Approximation and Projection,UMAP)进行文档聚类及可视化,根据复杂网络理论构建关键词共现网络,分析网络特性和节点联系,应用逼近理想解排序法(Technique for Order Preference by Similarity to an Ideal Solution,TOPSIS)获得节点重要性排序。研究结果显示:网络的平均聚类系数为0.71~0.965,覆盖了不同部门、岗位和环境,代表了民航应急管理的普遍情况;在人员兼职、空管单位、支线机场等应急体系建设存在不足的情况下,建议民航系统通过增设专职应急管理机构、优化资源配置、加强培训支持等方式提升应急管理能力。研究结果可为民航应急体系建设提供数据支持。
文摘中文电子病历命名实体识别主要是研究电子病历病程记录文书数据集,文章提出对医疗手术麻醉文书数据集进行命名实体识别的研究。利用轻量级来自Transformer的双向编码器表示(A Lite Bidirectional Encoder Representation from Transformers,ALBERT)预训练模型微调数据集和Tranfomers中的trainer训练器训练模型的方法,实现在医疗手术麻醉文书上识别手术麻醉事件命名实体与获取复杂麻醉医疗质量控制指标值。文章为医疗手术麻醉文书命名实体识别提供了可借鉴的思路,并且为计算复杂麻醉医疗质量控制指标值提供了一种新的解决方案。
文摘针对现有的数字化档案多标签分类方法存在分类标签之间缺少关联性的问题,提出一种用于档案多标签分类的深层神经网络模型ALBERT-Seq2Seq-Attention.该模型通过ALBERT(A Little BERT)预训练语言模型内部多层双向的Transfomer结构获取进行文本特征向量的提取,并获得上下文语义信息;将预训练提取的文本特征作为Seq2Seq-Attention(Sequence to Sequence-Attention)模型的输入序列,构建标签字典以获取多标签间的关联关系.将分类模型在3种数据集上分别进行对比实验,结果表明:模型分类的效果F1值均超过90%.该模型不仅能提高档案文本的多标签分类效果,也能关注标签之间的相关关系.
文摘Introduction: Seizures are one of the most common neurological complications in the infant period. The aim of our study was to describe the epidemiological, clinical, therapeutic and prognostic features of seizures in infants at the Albert Royer Children’s Hospital (Senegal). Materials and Methods: This was a retrospective, descriptive study from 1 January 2012 to 30 September 2018 of infants aged 0 days to 2 months who presented with seizures. Results: The hospital rate was 8.5%. Almost all the mothers (99.1%) had undergone at least 3 antenatal visits. Urogenital infection, gestational arterial hypertension and funicular anomalies were the main pregnancy-related pathologies. Delivery was vaginal in the majority of cases (80.9%). Most infants (43.6%) had not cried at birth. The majority of infants (63%) were born at term. Trophicity was normal in 68% of cases. The average age of the infants was 6.7 days. The main causes of seizures were hypoxic-ischemic encephalopathy (48.7%), metabolic disturbances (48.1%) and central ոеrvοսѕ system infections (15.6%). Phenobarbital was the 1st-line anticonvulsant. The case fatality rate was 39.5%. The main sequela observed were delayed psychomotor development (20.6%). Conclusion: Optimal management of infant seizures requires early diagnosis and etiological treatment by improving the quality of perinatal care to ensure better management of risk factors, as well as increasing the availability of neuroimaging equipment.
文摘园林植物知识图谱可为顾及区域适应性、观赏性和生态性等因子的绿化树种的选型提供知识支持。植物描述文本的实体识别及关系抽取是知识图谱构建的关键环节。针对植物领域未有公开的标注数据集,本文阐述了园林植物数据集的构建流程,定义了园林植物的概念体系结构,完成了园林植物语料库的构建。针对现有Word2vec、ELMo和BERT等语言模型存在无法解决多义词、融合上下文能力差、运行速度慢等缺点,提出了嵌入ALBERT(A Lite BERT)预训练语言模型的实体识别和关系抽取模型。ALBERT预训练的动态词向量能够有效地表示文本特征,将其分别输入到BiGRU-CRF命名实体识别模型和BiGRU-Attention关系抽取模型中进行训练,进一步提升实体识别和关系抽取的效果。在园林植物语料库上进行方法的有效性验证,结果表明ALBERT-BiGRU-CRF命名实体识别模型的F1值为0.9517,ALBERT-BiGRU-Attention关系抽取模型的F1值为0.9161,相较于经典的语言模型(如Word2vec、ELMo和BERT等)性能有较为显著的提升。因此基于ALBERT模型的实体与关系抽取任务能有效提高识别分类效果,可将其应用于植物描述文本的实体关系抽取任务中,为园林植物知识图谱自动构建提供方法。
