ID 原文 译文
58258 在真实数据集上的实验结果表明，所提出的方法获得了更好的情感分类准确率. Results of experiments conducted on real datasets show that the proposed approach achieves better sentiment classification accuracy than those state-of-the-art approaches.
58259 目前大部分的拼接检测算法主要检测图片是否经历了拼接篡改，而不是对拼接区域进行定位检测，对此，提出了一种基于改进 DeepLabv3 + 的拼接区域定位检测技术. As the existing splicing detection algorithms mainly detect whether the image has undergoneimage splicing，rather than regional localization of the spliced area.
58260 首先，改变原 DeepLabv3 + 网络的分类数; An improved DeepLabv3 + networkthat changes the categories of original DeepLabv3 + network is proposed to locate image regional forgery.
58261 其次，通过改造图像训练库，在训练库中加入含有人物的原图，对原图和篡改图的标签进行区别设置，引导改进的 DeepLabv3 +网络去学习原图人物和篡改人物特征的区别. In addition，the original images are added to the training dataset，and the ground truth of the originalimages are set to be black images，which will guide the network to learn the difference between the original images and the spliced images.
58262 实验结果显示，基于改进 DeepLabv3 + 的拼接区域定位检测技术在CASIA 数据库上取得了更好的拼接区域定位效果. Experimental results show that the improved DeepLabv3 + networkhas achieved a finer effect for locating image regional forgery than the existing algorithms on the CASIAdatabase.
58263 利用 Dempster 合成规则组合证据时，有可能会出现合成悖论，且其表现形式多样，很难直接获得它们之间的规律或关系. When applying Dempster’s rule of combination to fuse pieces of evidence，a combination paradox can occur. It is very difficult to acquire the law or relationship among these combination paradoxeswith some different manifestations.
58264 对此，利用规范分解法将证据的基本信任分配转换为一组广义简单支持函数的组合形式. In order to resolve this problem，the study is carried out by the aid ofcanonical decomposition which is used to transform the basic belief assignment into the combination of aset of generalized simple support function.
58265 从 2 个典型反例分析入手，得出它们都存在信任忽略这一共同问题，在此基础上进一步总结出了Ⅰ类和Ⅱ类 2 种信任忽略形式及其发生的规律. Firstly，two typical counterexamples are analyzed，and thenthe common characteristic between them，neglecting belief，is summarized. On the basis of this，a further analysis is made. Two forms of neglecting belief and the rules of causing them are generalized.
58266 对Ⅱ类信任忽略进一步分析，给出了不会发生Ⅱ类信任忽略的一般性结论. Afurther analysis of type Ⅱ neglecting belief is made. Some general conclusions，that type Ⅱ neglectingbelief cannot occur，are presented.
58267 为更合理地使用 Dempster 合成规则避免出现合成悖论提供了依据，也为更有效地改进证据组合方法提供了参考. These generalized rules and conclusions provide not only the basis forapplying Dempster’s rule more reasonably to avoid combination paradoxes，but also the reference formodifying the combination method more effectively.