Large scale automatic image annotation based on convolutional neural network

Automatic image annotation is one of the most important challenges in computer vision, which is critical to many real-world researches and applications. In this paper, we focus on the issue of large scale image annotation with deep learning. Firstly, considering the existing image data, especially the network images, most of the labels of themselves are inaccurate or imprecise. We propose a Multitask Voting (MV) method, which can improve the accuracy of original annotation to a certain extent, thereby enhancing the training effect of the model. Secondly, the MV method can also achieve the adaptive label, whereas most existing methods pre-specify the number of tags to be selected. Additionally, based on convolutional neural network, a large scale image annotation model MVAIACNN is constructed. Finally, we evaluate the performance with experiments on the MIRFlickr25K and NUS-WIDE datasets, and compare with other methods, demonstrating the effectiveness of the MVAIACNN.     

Research and Practice of Nondictionary Chinese Segmentation

ＲｅｓｅａｒｃｈａｎｄＰｒａｃｔｉｃｅｏｆＮｏｎｄｉｃｔｉｏｎａｒｙＣｈｉｎｅｓｅＳｅｇｍｅｎｔａｔｉｏｎＷＡＮＧＫａｉｚｈｕ；ＬＩＪｕｎｊｉｅ；ＷＵＹａｎ王开铸，李俊杰，吴岩（ＤｅｐａｒｔｍｅｎｔｏｆＣｏｍｐｕｔｅｒＳｃｉｅｎｃｅａｎｄＥｎｇｉｎｅ...     

The role of the corpus callosum in dichotic listening: A combined morphological and diffusion tensor imaging study.

The objective of the present study was to examine the role of the corpus callosum (CC) in dichotic listening. For this purpose, 40 right-handed healthy male participants were assessed with morphological and diffusion tensor imaging, which yielded macrostructural (midsagittal area) and microstructural (mean diffusion, fractional anisotropy) measures of the total CC and predefined subregions. Applying the standard consonant-vowel Bergen Dichotic Listening Test (including 3 different attentional instructions), the authors found dichotic listening performance to be substantially related to the integrity of the CC at both macro- and microstructural levels. Results indicate a dual role of the CC; it is relevant not only for the stimulus-driven (bottom-up) transfer of left-ear input to the left hemisphere but also for its attentional (top-down) modulation. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

基于长度的扩展方法的汉英句子对齐

本文提出了一种用于汉英平行语料库对齐的扩展方法。该扩展方法以基于长度的统计对齐方法为主,然后根据双语词典引入了词汇信息,而基于标点的方法作为对齐的后处理部分。这种扩展方法不仅避免了复杂的中文处理,例如,汉语分词和词性标注,而且在统计方法中引入了关键词信息,以提高句子对齐的正确率。本文中所用的双语语料是LDC 的关于香港的双语新闻报道。动态规划算法用于系统的实现。和单纯的基于长度的方法和词汇方法相比,我们的扩展方法提高了句子对齐的正确率,并且结果是比较理想的。     

基于深度学习的变电站硬压板状态检测与识别算法

针对变电站的一键顺控停、送电操作不当将导致投退操作的问题,提出了一种基于深度学习的变电站硬压板状态检测与识别算法.使用一个共享网络提取图像特征,基于多任务学习方法建立3个分支联合解决硬压板位置检测、投切状态检测和标识检测这3个任务;采集标注了8 000张硬压板图片数据用于训练和测试.结果表明,所提出的方法能够在提升硬压板状态识别精度的同时,也提升一键顺控操作的安全性.     

试议科技论文参考文献的标注与著录

参考文献是科技文的重要组成部分，是学术期刊引文进行数据统计和分析的重要信息源之一，通过对“CAJ－CD B／T1－1998中国学术期刊（光盘版）检索与评价数据规范（试行）”的执行，针对目前科技论文来稿的诸多存在问题进行了评议。为国家相关规范的进一步实施与贯彻作一些基础工作，以期更好地提高期刊的质量和水平。     

北京大学现代汉语语料库基本加工规范(续)

北京大学计算语言学研究所已经完成了一个有2700万汉字的现代汉语语料库的基本加工。加工项目除词语切分和词性标注外,还包括专有名词(人名、地名、团体机构名称等)标注、语素子类标注以及动词、形容词的特殊用法标注。这项大规模语言工程的顺利完成得益于事先制订并不断完善的规范。发表《北京大学现代汉语语料库基本加工规范》是为了抛砖引玉,更广泛地向专家、同行征询意见,以便进一步修订。     

Hemispheric interactions are different in left-handed individuals.

In a previous study, N. Cherbuin and C. Brinkman (2006) showed that in right-handed participants, interhemispheric transfer time (measured with A. T. Poffenberger's, 1912, paradigm) was a significant predictor of the efficiency of hemispheric interactions (measured with a split visual field, letter-matching task). No effect was found for degree of handedness in this study. This was surprising because handedness has been shown to be associated with differences in the morphology and the structure of the corpus callosum, and cerebral anatomical lateralization, as well as functional lateralization both in behavioral and scanning studies. Because these findings were found in a large sample, but one limited to right-handed participants, the aim of the present study was to determine whether a similar relationship was present between interhemispheric transfer time and hemispheric interaction in left-handed participants (using identical measures) and to assess whether the analysis of a larger sample that comprised both left- and right-handed participants might reveal an effect of handedness. Results demonstrate significant handedness effects, suggesting that left-handed individuals tend to have more efficient hemispheric interactions. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

语料库中熟语的标记问题

熟语是自然语言中普遍存在的语言现象。本文分析了国内现有语料库对熟语的标注方式,发现这种方式对语料库的进一步加工是有问题的。为了在语料库标注阶段把熟语问题处理好,本文从信息处理的角度将熟语中的成语、惯用语、歇后语、习用语、专门语以及缩略语归为固定语的范畴,进而提出根据固定语的语法功能给定词性标记,再根据它们的词汇特征给定词汇范畴标记的双层标记法,这样在一定程度上解决了熟语的语料库标注问题。     

汉语词性标注排歧方法探讨

该文将概率统计的二元模型与三元模型用于汉语词性自动标注,在算法为线性阶的时间复杂度的情况下,对20万训练集和1万的测试集,分别进行封闭测试和开放测试,对稀疏矩阵零元素及词性标注的结果做了统计分析。