基于BERTCA的新闻实体与正文语义相关度计算模型

目前的搜索引擎仍然存在“重形式,轻语义”的问题,无法做到对搜索关键词和文本的深层次语义理解,因此语义检索成为当前搜索引擎中亟需解决的问题。为了提高搜索引擎的语义理解能力,该文提出一种语义相关度的计算方法。首先,标注了金融类新闻标题实体与新闻正文语义相关度语料1万条,然后建立新闻实体与正文语义相关度计算的BERTCA(Bidirectional Encoder Representation from Transformers Co-Attention)模型,通过使用BERT预训练模型,综合考虑细粒度的实体和粗粒度的正文的语义信息,然后经过协同注意力,实现实体与正文的语义匹配,不仅能计算出金融新闻实体与新闻正文之间的相关度,还能根据相关度阈值来判定相关度类别,实验表明该模型在1万条标注语料上准确率超过95%,优于目前主流模型,最后通过具体搜索示例展示了该模型的优秀性能。     

Human DDX17 Unwinds Rift Valley Fever Virus Non-Coding RNAs

Rift Valley fever virus (RVFV) is a mosquito-transmitted virus from the Bunyaviridae family that causes high rates of mortality and morbidity in humans and ruminant animals. Previous studies indicated that DEAD-box helicase 17 (DDX17) restricts RVFV replication by recognizing two primary non-coding RNAs in the S-segment of the genome: the intergenic region (IGR) and 5′ non-coding region (NCR). However, we lack molecular insights into the direct binding of DDX17 with RVFV non-coding RNAs and information on the unwinding of both non-coding RNAs by DDX17. Therefore, we performed an extensive biophysical analysis of the DDX17 helicase domain (DDX17_135–555) and RVFV non-coding RNAs, IGR and 5’ NCR. The homogeneity studies using analytical ultracentrifugation indicated that DDX17_135–555, IGR, and 5’ NCR are pure. Next, we performed small-angle X-ray scattering (SAXS) experiments, which suggested that DDX17 and both RNAs are homogenous as well. SAXS analysis also demonstrated that DDX17 is globular to an extent, whereas the RNAs adopt an extended conformation in solution. Subsequently, microscale thermophoresis (MST) experiments were performed to investigate the direct binding of DDX17 to the non-coding RNAs. The MST experiments demonstrated that DDX17 binds with the IGR and 5’ NCR with a dissociation constant of 5.77 ± 0.15 µM and 9.85 ± 0.11 µM, respectively. As DDX17_135–555 is an RNA helicase, we next determined if it could unwind IGR and NCR. We developed a helicase assay using MST and fluorescently-labeled oligos, which suggested DDX17_135–555 can unwind both RNAs. Overall, our study provides direct evidence of DDX17_135–555 interacting with and unwinding RVFV non-coding regions.     

Mechanistic Studies on Trichoacorenol Synthase from Amycolatopsis benzoatilytica

Isotopic labeling experiments performed with a newly identified bacterial trichoacorenol synthase established a 1,5-hydride shift occurring in the cyclization mechanism. During EI-MS analysis, major fragments of the sesquiterpenoid were shown to arise via cryptic hydrogen movements. Therefore, the interpretation of earlier results regarding the cyclization mechanism obtained by feeding experiments in Trichoderma is revised.     

中国传统白酒中微量成分的来源与作用

根据中国传统白酒微量成分研究取得的新成果,系统地总结了微量成分来源及其作用。主要来源有3条途径:发酵过程中微生物和酶的催化反应、贮存过程中的化学变化和生产原料特殊成分的化学反应,主要起到了呈香呈味、成格成型和生理功能三方面的作用。生产原料对中国传统白酒微量成分的影响、高产白酒主体风味成分的基因菌构建与应用和白酒功能因子的深入研究等方向是未来研究的重点。     

基于卷积神经网络的单目深度估计

针对利用深度学习方法对街道图像进行深度估计，提出采用语义分割的方法解决深度图出现边界模糊等问题；估计深度通过左右视角图生成视差图进行无监督的训练。在网络模型中添加语义分割层，采取多个空洞卷积并行的结构增加感受野，同时减少了图像下采样的次数，降低了由于下采样带来的信息损失，使得的结果更加准确。这也是在深度估计中首次与空洞卷积相结合增加准确率。通过对KITTI街道数据集进行训练，与现有结果相比，除了增加检测准确性，降低错误率之外，使得效果图中的物体更加清晰，并且在效果图中还保留了一些原模型中被忽视掉的细节信息，将原始图像更加完整的表现出来。     

面向连续查询的敏感语义位置隐私保护方案

针对连续查询位置服务中构造匿名区域未考虑语义位置信息导致敏感隐私泄露问题，通过设计[(K,θ)]-隐私模型，提出一种路网环境下面向连续查询的敏感语义位置隐私保护方案。该方案利用Voronoi图将城市路网预先划分为独立的Voronoi单元，依据用户的移动路径和移动速度，选择具有相似特性的其他[K-1]个用户，构建匿名用户集；利用匿名用户集用户设定的敏感语义位置类型和语义安全阈值，以及用户所处语义位置的Voronoi单元，构建满足[(K,θ)]-隐私模型的语义安全匿名区域，可以同时防止连续查询追踪攻击和语义推断攻击。实验结果表明，与SCPA算法相比，该方案在隐私保护程度上提升约15%，系统开销上降低约20%。     

新加坡食品较健康选择标志系统经验启示

文章介绍了新加坡较健康选择标志系统基本情况,从规格与使用要求、申请程序、适用范围等进行阐述,指出新加坡作为第一个实施食品FOP标签系统的亚洲国家,其经验做法对中国乃至多数亚洲国家有重要借鉴意义,中国FOP标签系统应具有充分的依据,严格规范申请和使用行为并适度扩大应用范围。     

R2D2: A scalable deep learning toolkit for medical imaging segmentation

Deep learning has gained a significant popularity in recent years thanks to its tremendous success across a wide range of relevant fields of applications, including medical image analysis domain in particular. Although convolutional neural networks (CNNs) based medical applications have been providing powerful solutions and revolutionizing medicine, efficiently training of CNNs models is a tedious and challenging task. It is a computationally intensive process taking long time and rare system resources, which represents a significant hindrance to scientific research progress. In order to address this challenge, we propose in this article, R2D2, a scalable intuitive deep learning toolkit for medical imaging semantic segmentation. To the best of our knowledge, the present work is the first that aims to tackle this issue by offering a novel distributed versions of two well-known and widely used CNN segmentation architectures [ie, fully convolutional network (FCN) and U-Net]. We introduce the design and the core building blocks of R2D2. We further present and analyze its experimental evaluation results on two different concrete medical imaging segmentation use cases. R2D2 achieves up to 17.5× and 10.4× speedup than single-node based training of U-Net and FCN, respectively, with a negligible, though still unexpected segmentation accuracy loss. R2D2 offers not only an empirical evidence and investigates in-depth the latest published works but also it facilitates and significantly reduces the effort required by researchers to quickly prototype and easily discover cutting-edge CNN configurations and architectures.     

海参硫酸软骨素的荧光标记方法

海参硫酸软骨素是海参体壁的重要组成物质之一,已被证实具有多种生理调节功能,在功能食品开发方面展现出良好的应用前景。与其他碳水化合物类似,海参硫酸软骨素缺乏特异的发色及发光基团。基于此背景,本研究旨在对海参硫酸软骨素进行荧光标记。以6-氨基荧光素为荧光染料并配合溴化氢活化法成功对该多糖进行了荧光衍生,每分子标记产物中约含4.4分子荧光染料,其最佳荧光激发波长为495 nm、发射波长为520 nm。分子质量测定、硫酸根质量分数测定及傅里叶变换红外光谱比较分析的结果表明,标记前后多糖的结构特征未发生明显变化,标记产物在中性及碱性环境中展现出较强的荧光强度及稳定性,溶液环境中的NaCl对标记产物的影响有限。海参硫酸软骨素荧光标记方法的建立为后续研究该多糖的吸收代谢过程、功能机理、生物分子间相互作用等科学问题提供了有利条件。     

伪标签置信选择的半监督集成学习视频语义检测

在视频语义检测中，有标记样本不足会严重影响检测的性能，而且伪标签样本中的噪声也会导致集成学习基分类器性能提升不足。为此，提出一种伪标签置信选择的半监督集成学习算法。首先，在三个不同的特征空间上训练出三个基分类器，得到基分类器的标签矢量；然后，引入加权融合样本所属某个类别的最大概率与次大概率的误差和样本所属某个类别的最大概率与样本所属其他各类别的平均概率的误差，作为基分类器的标签置信度，并融合标签矢量和标签置信度得到样本的伪标签和集成置信度；接着，选择集成置信度高的样本加入到有标签的样本集，迭代训练基分类器；最后，采用训练好的基分类器集成协作检测视频语义概念。该算法在实验数据集UCF11上的平均准确率到达了83.48%，与Co-KNN-SVM算法相比，平均准确率提高了3.48个百分点。该算法选择的伪标签能体现样本所属类别与其他类别的总体差异性，又能体现所属类别的唯一性，可减少利用伪标签样本的风险，有效提高视频语义概念检测的准确率。