Linguistic studies of context benefit translation a lot,as context is first accepted as a social linguistic concept in its long history of development.Obviously,context has become increasingly crucial for translation ...Linguistic studies of context benefit translation a lot,as context is first accepted as a social linguistic concept in its long history of development.Obviously,context has become increasingly crucial for translation and translation studies.Some scholars even suggest that an independent contextual vision of translation be established to distinguish context of translation from those of text linguistics,pragmatics and cognitive linguistics,etc.It is a set of cohesive devices that constitute the coherence of a discourse.Some linguist classifies the devices into two groups:internal cohesive devices and external cohesive devices.Absolutely,each coherent discourse is contextually adapted and the thesis attempts to transplant context of translation into the soil of discourse,so that contextual elements can be categorized accordingly and translation process will be considered as reconstruction of textual coherence in the target discourse under the constraint of context.Therefore the translators should enhance their contextual awareness and internalize contextual knowledge into their minds in a bid for an effective translation.展开更多
To be different from traditional algorithms for concept lattice constructing, a method based on nth-order context kernel is suggested in this paper. The context kernels support generating small lattices for sub-contex...To be different from traditional algorithms for concept lattice constructing, a method based on nth-order context kernel is suggested in this paper. The context kernels support generating small lattices for sub-contexts split by a given context. The final concept lattice is reconstructed by combining these small lattices. All relevant algorithms are implemented in a system IsoFCA. Test shows that the method yields concept lattices in lower time complexity than Godin algorithm in practical case.展开更多
针对三维重建对细小特征及边缘区域重建欠佳的问题,提出了一个基于特征对齐与上下文引导的多视图三维重建网络,即AGA-MVSNet。首先,构建了一个特征对齐模块(FA)与特征选择模块(FS),能够将特征金字塔不同层级的特征先对齐之后再进行融合...针对三维重建对细小特征及边缘区域重建欠佳的问题,提出了一个基于特征对齐与上下文引导的多视图三维重建网络,即AGA-MVSNet。首先,构建了一个特征对齐模块(FA)与特征选择模块(FS),能够将特征金字塔不同层级的特征先对齐之后再进行融合,提高对小尺寸物体和边缘区域的特征提取能力;然后,在代价体正则化中加入了一个上下文引导模块,该模块能够在略微增加运行内存的情况下充分利用周围信息,增强成本体积之间的相关性,提高三维重建的精度与完整度;最后,在DTU数据集上进行了实验,实验结果表明,该方法相比于基准网络CasMVSNet精度提升了2.2%,整体重建质量提升了2.5%。此外,在Tanks and Temples数据集上的表现相较一些已知的方法也十分优异,且在BlendedMVS数据集上也生成了不错的点云效果。展开更多
针对现有单图像超分辨率方法在重建过程中容易忽略原图像中不同结构-纹理的差异与联系,导致生成的高分辨率图像缺乏纹理细节并存在伪影的问题,提出了纹理细节恢复的图像超分辨率重建算法。该方法由梯度分支、纹理分支和图像超分辨率分...针对现有单图像超分辨率方法在重建过程中容易忽略原图像中不同结构-纹理的差异与联系,导致生成的高分辨率图像缺乏纹理细节并存在伪影的问题,提出了纹理细节恢复的图像超分辨率重建算法。该方法由梯度分支、纹理分支和图像超分辨率分支组成。其中,在梯度分支和纹理分支之间使用了类注意力模块处理二者的特征混淆问题,并通过双向特征融合模块实现了对结构特征与纹理特征的相互促进,作为先验信息以达到纹理细节信息增强的目的。此外,在图像超分辨率分支还通过构建特征恢复模块,利用浅层和深层信息帮助网络保留了图像中更丰富的上下文信息和纹理细节。该方法通过在DIV2K数据集上进行了网络训练,并在5个基准测试集Set5、Set14、BSD100、Urban100和MANGA109上进行了实验,峰值信噪比(PSNR, Peak Signal to Noise Ratio)分别:37.88dB、33.28dB、32.0781dB、31.89dB、38.39dB,相比现有方法均有显著提升。实验结果表明,本文方法获得了有效的重建图像并且保留更多的图像细节,生成具有边缘清晰和逼真细节的超分辨率图像。展开更多
文摘Linguistic studies of context benefit translation a lot,as context is first accepted as a social linguistic concept in its long history of development.Obviously,context has become increasingly crucial for translation and translation studies.Some scholars even suggest that an independent contextual vision of translation be established to distinguish context of translation from those of text linguistics,pragmatics and cognitive linguistics,etc.It is a set of cohesive devices that constitute the coherence of a discourse.Some linguist classifies the devices into two groups:internal cohesive devices and external cohesive devices.Absolutely,each coherent discourse is contextually adapted and the thesis attempts to transplant context of translation into the soil of discourse,so that contextual elements can be categorized accordingly and translation process will be considered as reconstruction of textual coherence in the target discourse under the constraint of context.Therefore the translators should enhance their contextual awareness and internalize contextual knowledge into their minds in a bid for an effective translation.
基金Supported by the National Natural Science Foun-dation of China (60275022) the Natural Science Foundation ofHenan Province (0311011700)
文摘To be different from traditional algorithms for concept lattice constructing, a method based on nth-order context kernel is suggested in this paper. The context kernels support generating small lattices for sub-contexts split by a given context. The final concept lattice is reconstructed by combining these small lattices. All relevant algorithms are implemented in a system IsoFCA. Test shows that the method yields concept lattices in lower time complexity than Godin algorithm in practical case.
文摘针对三维重建对细小特征及边缘区域重建欠佳的问题,提出了一个基于特征对齐与上下文引导的多视图三维重建网络,即AGA-MVSNet。首先,构建了一个特征对齐模块(FA)与特征选择模块(FS),能够将特征金字塔不同层级的特征先对齐之后再进行融合,提高对小尺寸物体和边缘区域的特征提取能力;然后,在代价体正则化中加入了一个上下文引导模块,该模块能够在略微增加运行内存的情况下充分利用周围信息,增强成本体积之间的相关性,提高三维重建的精度与完整度;最后,在DTU数据集上进行了实验,实验结果表明,该方法相比于基准网络CasMVSNet精度提升了2.2%,整体重建质量提升了2.5%。此外,在Tanks and Temples数据集上的表现相较一些已知的方法也十分优异,且在BlendedMVS数据集上也生成了不错的点云效果。
文摘针对现有单图像超分辨率方法在重建过程中容易忽略原图像中不同结构-纹理的差异与联系,导致生成的高分辨率图像缺乏纹理细节并存在伪影的问题,提出了纹理细节恢复的图像超分辨率重建算法。该方法由梯度分支、纹理分支和图像超分辨率分支组成。其中,在梯度分支和纹理分支之间使用了类注意力模块处理二者的特征混淆问题,并通过双向特征融合模块实现了对结构特征与纹理特征的相互促进,作为先验信息以达到纹理细节信息增强的目的。此外,在图像超分辨率分支还通过构建特征恢复模块,利用浅层和深层信息帮助网络保留了图像中更丰富的上下文信息和纹理细节。该方法通过在DIV2K数据集上进行了网络训练,并在5个基准测试集Set5、Set14、BSD100、Urban100和MANGA109上进行了实验,峰值信噪比(PSNR, Peak Signal to Noise Ratio)分别:37.88dB、33.28dB、32.0781dB、31.89dB、38.39dB,相比现有方法均有显著提升。实验结果表明,本文方法获得了有效的重建图像并且保留更多的图像细节,生成具有边缘清晰和逼真细节的超分辨率图像。
