Cu x S thin films are implanted by low energy N + ion beam. The influences of the energy and dose of N + ion implantation on Cu x S films are investigated. The results show that the ratio of copper to sulfur is increa...Cu x S thin films are implanted by low energy N + ion beam. The influences of the energy and dose of N + ion implantation on Cu x S films are investigated. The results show that the ratio of copper to sulfur is increased to some extent, the constituents of the film are turned to rich copper phase from rich sulfurous phase after ion beam irradiation. X-ray diffraction spectrum and optical transmission spectra of sample have confirmed the results.展开更多
The Basofil fiber is a new type of high-temperature fiber. Its Limiting Oxygen Index (LOI) value reaches 32. It does not melt, drop down or smolder when it contacts fire. The fiber is condensed from melamine and forma...The Basofil fiber is a new type of high-temperature fiber. Its Limiting Oxygen Index (LOI) value reaches 32. It does not melt, drop down or smolder when it contacts fire. The fiber is condensed from melamine and formaldehyde into a cross-linked with methylene-ether and methylene bridge bond. This thesis discusses the change of fiber structure and performance after the acid treatment.展开更多
现有的中文事件抽取方法存在触发词和论元依赖建模不足的问题,削弱事件内的信息交互,导致论元抽取性能低下,特别是论元角色存在重叠的情况下.对此,文中提出基于图注意力和表指针网络的中文事件抽取方法(Chinese Event Extraction Method...现有的中文事件抽取方法存在触发词和论元依赖建模不足的问题,削弱事件内的信息交互,导致论元抽取性能低下,特别是论元角色存在重叠的情况下.对此,文中提出基于图注意力和表指针网络的中文事件抽取方法(Chinese Event Extraction Method Based on Graph Attention and Table Pointer Network,ATCEE).首先,融合预训练字符向量和词性标注向量作为特征输入,并利用双向长短期记忆网络,得到事件文本的强化语义特征.再将字符级建模的依存句法图引入图注意力网络,捕获文本中各组成成分的长距离依赖关系.然后,使用表填充的方法进行特征融合,进一步增强触发词和其对应的所有论元之间的依赖性.最后,将学习得到的表特征输入全连接层和表指针网络层,进行触发词和论元的联合抽取,使用表指针网络对论元边界进行解码,更好地识别长论元实体.实验表明:ATCEE在ACE2005和DuEE1.0这两个中文基准数据集上都有明显的性能提升,并且字符级依存特征和表填充策略在一定程度上可以解决论元角色重叠问题.ATCEE源代码地址如下:https://github.com/event6/ATCEE.展开更多
文摘Cu x S thin films are implanted by low energy N + ion beam. The influences of the energy and dose of N + ion implantation on Cu x S films are investigated. The results show that the ratio of copper to sulfur is increased to some extent, the constituents of the film are turned to rich copper phase from rich sulfurous phase after ion beam irradiation. X-ray diffraction spectrum and optical transmission spectra of sample have confirmed the results.
文摘The Basofil fiber is a new type of high-temperature fiber. Its Limiting Oxygen Index (LOI) value reaches 32. It does not melt, drop down or smolder when it contacts fire. The fiber is condensed from melamine and formaldehyde into a cross-linked with methylene-ether and methylene bridge bond. This thesis discusses the change of fiber structure and performance after the acid treatment.
基金Supported by National High Technology Research and Development Program of China （863 Program） （2006AA04Z182）, National Creative Research Groups Science Foundation of China （60721062）, and National Natural Science Foundation of China （60736021）
文摘现有的中文事件抽取方法存在触发词和论元依赖建模不足的问题,削弱事件内的信息交互,导致论元抽取性能低下,特别是论元角色存在重叠的情况下.对此,文中提出基于图注意力和表指针网络的中文事件抽取方法(Chinese Event Extraction Method Based on Graph Attention and Table Pointer Network,ATCEE).首先,融合预训练字符向量和词性标注向量作为特征输入,并利用双向长短期记忆网络,得到事件文本的强化语义特征.再将字符级建模的依存句法图引入图注意力网络,捕获文本中各组成成分的长距离依赖关系.然后,使用表填充的方法进行特征融合,进一步增强触发词和其对应的所有论元之间的依赖性.最后,将学习得到的表特征输入全连接层和表指针网络层,进行触发词和论元的联合抽取,使用表指针网络对论元边界进行解码,更好地识别长论元实体.实验表明:ATCEE在ACE2005和DuEE1.0这两个中文基准数据集上都有明显的性能提升,并且字符级依存特征和表填充策略在一定程度上可以解决论元角色重叠问题.ATCEE源代码地址如下:https://github.com/event6/ATCEE.
