BACKGROUND Esophageal adenocarcinoma(EAC)derived from long-segment Barrett’s esophagus(LSBE)is extremely rare in Asia.LSBE-related EAC is often difficult to diagnose in the horizontal extent.If the tumor has spread t...BACKGROUND Esophageal adenocarcinoma(EAC)derived from long-segment Barrett’s esophagus(LSBE)is extremely rare in Asia.LSBE-related EAC is often difficult to diagnose in the horizontal extent.If the tumor has spread throughout the LSBE,whole circumferential endoscopic submucosal dissection(ESD)should be performed,which is difficult to complete safely.Additionally,whole circumferential ESD can bring refractory postoperative stenosis.We hereby report a case of EAC involving the whole circumference of the LSBE,achieving complete endoscopic removal without complications.CASE SUMMARY An 85-year-old man with the chief complaint of dysphagia underwent esophagogastroduodenoscopy.We suspected a flat-type cancerous lesion that extended the whole circumference of the LSBE(C 3.5,M 4.0)using narrow-band imaging magnification endoscopy(NBI-M).We achieved circumferential en bloc resection of the lesion safely with special ESD techniques.Histology of the ESD specimens demonstrated that the superficial EAC extended the whole circumference of the LSBE,and papillary or well-differentiated tubular adenocarcinoma was confined in the lamina propria mucosa showing a vertical negative margin.To prevent post-ESD stenosis,we performed endoscopic local injection of steroids,followed by oral administration of steroids.There was no evidence of esophageal refractory stenosis or tumor recurrence 30 mo after ESD.In summary,we experienced a rare case of LSBE-related EAC.The horizontal tumor extent was accurately diagnosed by NBI-M.Additionally,we achieve whole circumferential ESD safely without postoperative refractory stenosis.CONCLUSION NBI-M,ESD,and steroid therapy enabled the curative resection of superficial full circumferential LSBE-related EAC without refractory postoperative stenosis.展开更多
Objective To investigate the operative method and evaluate the clinical outcomes of long proximal femoral nail anti-rotation (PFNA-long) in treating long-segment fracture in middle-up part of femoral bone.Methods From...Objective To investigate the operative method and evaluate the clinical outcomes of long proximal femoral nail anti-rotation (PFNA-long) in treating long-segment fracture in middle-up part of femoral bone.Methods From June 2006 to展开更多
针对畜禽疫病文本语料匮乏、文本内包含大量疫病名称及短语等未登录词问题,提出了一种结合词典匹配的BERT-BiLSTM-CRF畜禽疫病文本分词模型。以羊疫病为研究对象,构建了常见疫病文本数据集,将其与通用语料PKU结合,利用BERT(Bidirectiona...针对畜禽疫病文本语料匮乏、文本内包含大量疫病名称及短语等未登录词问题,提出了一种结合词典匹配的BERT-BiLSTM-CRF畜禽疫病文本分词模型。以羊疫病为研究对象,构建了常见疫病文本数据集,将其与通用语料PKU结合,利用BERT(Bidirectional encoder representation from transformers)预训练语言模型进行文本向量化表示;通过双向长短时记忆网络(Bidirectional long short-term memory network,BiLSTM)获取上下文语义特征;由条件随机场(Conditional random field,CRF)输出全局最优标签序列。基于此,在CRF层后加入畜禽疫病领域词典进行分词匹配修正,减少在分词过程中出现的疫病名称及短语等造成的歧义切分,进一步提高了分词准确率。实验结果表明,结合词典匹配的BERT-BiLSTM-CRF模型在羊常见疫病文本数据集上的F1值为96.38%,与jieba分词器、BiLSTM-Softmax模型、BiLSTM-CRF模型、未结合词典匹配的本文模型相比,分别提升11.01、10.62、8.3、0.72个百分点,验证了方法的有效性。与单一语料相比,通用语料PKU和羊常见疫病文本数据集结合的混合语料,能够同时对畜禽疫病专业术语及疫病文本中常用词进行准确切分,在通用语料及疫病文本数据集上F1值都达到95%以上,具有较好的模型泛化能力。该方法可用于畜禽疫病文本分词。展开更多
[目的/意义]本文提出基于长短时记忆(Long short-term memory,LSTM)神经网络和条件随机场(Conditional Random Field,CRF)的藏文分词模型。[方法/过程]引入注意力机制,获取更多特征信息,提升模型关注上下文信息与当前音节之间联系;提出...[目的/意义]本文提出基于长短时记忆(Long short-term memory,LSTM)神经网络和条件随机场(Conditional Random Field,CRF)的藏文分词模型。[方法/过程]引入注意力机制,获取更多特征信息,提升模型关注上下文信息与当前音节之间联系;提出一种音节扩展方法,获取更多的输入特征信息与语料信息,增强模型单音节特征信息以获取更多语义信息的能力。[局限]本文在西藏大学数据集12261条的基础上,扩充至74384条,形成Tibetan-News数据集。[结果/结论]实验结果表明,在模型中加入注意力机制并使用音节扩展方法后,模型在Tibetan-News数据集上的精确率、召回率和F1分别提升2.9%、3.5%和3.2%。基于本文模型的分词系统已在工程上应用推广。展开更多
文摘BACKGROUND Esophageal adenocarcinoma(EAC)derived from long-segment Barrett’s esophagus(LSBE)is extremely rare in Asia.LSBE-related EAC is often difficult to diagnose in the horizontal extent.If the tumor has spread throughout the LSBE,whole circumferential endoscopic submucosal dissection(ESD)should be performed,which is difficult to complete safely.Additionally,whole circumferential ESD can bring refractory postoperative stenosis.We hereby report a case of EAC involving the whole circumference of the LSBE,achieving complete endoscopic removal without complications.CASE SUMMARY An 85-year-old man with the chief complaint of dysphagia underwent esophagogastroduodenoscopy.We suspected a flat-type cancerous lesion that extended the whole circumference of the LSBE(C 3.5,M 4.0)using narrow-band imaging magnification endoscopy(NBI-M).We achieved circumferential en bloc resection of the lesion safely with special ESD techniques.Histology of the ESD specimens demonstrated that the superficial EAC extended the whole circumference of the LSBE,and papillary or well-differentiated tubular adenocarcinoma was confined in the lamina propria mucosa showing a vertical negative margin.To prevent post-ESD stenosis,we performed endoscopic local injection of steroids,followed by oral administration of steroids.There was no evidence of esophageal refractory stenosis or tumor recurrence 30 mo after ESD.In summary,we experienced a rare case of LSBE-related EAC.The horizontal tumor extent was accurately diagnosed by NBI-M.Additionally,we achieve whole circumferential ESD safely without postoperative refractory stenosis.CONCLUSION NBI-M,ESD,and steroid therapy enabled the curative resection of superficial full circumferential LSBE-related EAC without refractory postoperative stenosis.
文摘Objective To investigate the operative method and evaluate the clinical outcomes of long proximal femoral nail anti-rotation (PFNA-long) in treating long-segment fracture in middle-up part of femoral bone.Methods From June 2006 to
文摘针对畜禽疫病文本语料匮乏、文本内包含大量疫病名称及短语等未登录词问题,提出了一种结合词典匹配的BERT-BiLSTM-CRF畜禽疫病文本分词模型。以羊疫病为研究对象,构建了常见疫病文本数据集,将其与通用语料PKU结合,利用BERT(Bidirectional encoder representation from transformers)预训练语言模型进行文本向量化表示;通过双向长短时记忆网络(Bidirectional long short-term memory network,BiLSTM)获取上下文语义特征;由条件随机场(Conditional random field,CRF)输出全局最优标签序列。基于此,在CRF层后加入畜禽疫病领域词典进行分词匹配修正,减少在分词过程中出现的疫病名称及短语等造成的歧义切分,进一步提高了分词准确率。实验结果表明,结合词典匹配的BERT-BiLSTM-CRF模型在羊常见疫病文本数据集上的F1值为96.38%,与jieba分词器、BiLSTM-Softmax模型、BiLSTM-CRF模型、未结合词典匹配的本文模型相比,分别提升11.01、10.62、8.3、0.72个百分点,验证了方法的有效性。与单一语料相比,通用语料PKU和羊常见疫病文本数据集结合的混合语料,能够同时对畜禽疫病专业术语及疫病文本中常用词进行准确切分,在通用语料及疫病文本数据集上F1值都达到95%以上,具有较好的模型泛化能力。该方法可用于畜禽疫病文本分词。
文摘[目的/意义]本文提出基于长短时记忆(Long short-term memory,LSTM)神经网络和条件随机场(Conditional Random Field,CRF)的藏文分词模型。[方法/过程]引入注意力机制,获取更多特征信息,提升模型关注上下文信息与当前音节之间联系;提出一种音节扩展方法,获取更多的输入特征信息与语料信息,增强模型单音节特征信息以获取更多语义信息的能力。[局限]本文在西藏大学数据集12261条的基础上,扩充至74384条,形成Tibetan-News数据集。[结果/结论]实验结果表明,在模型中加入注意力机制并使用音节扩展方法后,模型在Tibetan-News数据集上的精确率、召回率和F1分别提升2.9%、3.5%和3.2%。基于本文模型的分词系统已在工程上应用推广。
