基于桥连接的词典学习方法的语义解析

现阶段的语义解析方法大部分都基于组合语义,这类方法的核心就是词典。词典是词汇的集合,词汇定义了自然语言句子中词语到知识库本体中谓词的映射。语义解析一直面临着词典中词汇覆盖度不够的问题。针对此问题,该文在现有工作的基础上,提出了基于桥连接的词典学习方法,该方法能够在训练中自动引入新的词汇并加以学习,为了进一步提高新学习到的词汇的准确度,该文设计了新的词语—二元谓词的特征模板,并使用基于投票机制的核心词典获取方法。该文在两个公开数据集(WebQuestions和Free917)上进行了对比实验,实验结果表明,该文方法能够学习到新的词汇,提高词汇的覆盖度,进而提升语义解析系统的性能,特别是召回率。     

弱相互作用调控表面活性剂自组装(Ⅱ)——表面活性剂的结构与设计

介绍了表面活性剂的结构、性质以及分类,综述了阴离子表面活性剂、阳离子表面活性剂、两性离子表面活性剂、非离子表面活性剂的分子间弱相互作用。另外,简述了低聚型表面活性剂和其他新型表面活性剂的结构及应用。     

基于SCAE⁃ACGAN的直升机行星齿轮裂纹故障诊断

直升机行星传动轮系结构复杂、工况多变,其振动信号受工况影响大,在故障样本较少的情况下导致行星齿轮箱故障诊断准确率不高,早期故障诊断困难。针对上述问题,提出将堆栈收缩自动编码网络(stacked contractive autoencoder,简称SCAE)与辅助分类生成式对抗网络(auxiliary classifier generative adversarial networks,简称ACGAN)相结合的SCAE-ACGAN故障诊断方法。ACGAN的生成器产生与真实样本具有类似分布的生成样本,扩展训练样本集,并与真实样本一起输入至判别器进行训练。ACGAN采用SCAE作为判别器,利用SCAE良好的抗数据波动能力,从扩展样本集中挖掘出有效的深度特征,并实现样本的真伪与类别的判定。ACGAN的判别器和生成器在对抗学习训练机制下交替优化,提高方法的样本生成质量与故障判定能力。将SCAE-ACGAN应用于直升机行星轮裂纹故障诊断,结果表明,SCAE-ACGAN的故障诊断性能好,在样本数量少与工况变化情况下具有较好的健壮性和适应性。     

紫色马铃薯酒发酵工艺优化及主发酵期花青素与色泽的动态变化

以紫色马铃薯为原料，在单因素的基础上，通过响应面法对紫色马铃薯酒的发酵工艺参数进行优化。最佳发酵工艺参数为:马铃薯与面粉比1:0.6、料水比1:1 (g/mL)、α-淀粉酶添加量8.5 U/g、酒曲添加量0.45%。此工艺在28 ℃下发酵14 d，得到的紫色马铃薯酒酒精度13.5%vol，花青素含量166.34 mg/mL，酒体澄清透明呈紫红色，酸甜适中，有独特酒香，综合感官评定得分90分。测定主发酵期花青素和色泽的动态变化，分析表明，紫色马铃薯酒在发酵过程中花青素呈现先上升后下降的趋势，色度呈现先下降后趋于平缓的趋势，色调则与之相反。聚合色度和褐变指数总体均呈现逐渐上升的趋势。试验结果为工业化生产紫色马铃薯酒提供理论依据。     

Low-temperature synthesis of SiC nanowires with Ni catalyst

SiC nano wires were fabricated on the silicon substrate dipped with a layer of Ni catalyst at 900 ℃ by gas pressure annealing processing. The morphologies and crystal structures were determined by scanning electron microscopy(SEM), transmission electron microscopy(TEM)and X-ray diffraction(XRD). The results show that the assynthesized nanowires are β-SiC single crystalline with diameter range of 50-100 nm, and length of tens of micron by directly annealing at 900 ℃. The SiC nano wires grow along the [111] direction with highly uniform morphology. And the possible growth mechanism of SiC nano wires is proposed.The present work provides an efficient strategy for the production of high-quality SiC nano wires.     

Different hydrothermal corrosion results between two sides of sodalime float glass and the possible corrosion mechanism

Sodalime float (SF) glass is widely used in our societies and industries. Hydrothermal corrosion method is one of the effective ways to prepare a superhydrophobic glass, but there is still lack of knowledge about hydrothermal corrosion behavior and mechanism of SF glass. We have hydrothermally treated SF glass at 180 °C for different time, and tried to reveal the aqueous corrosion process of SF glass. We have characterized the morphologies and chemical compositions of samples, and found that (1) the two sides of SF glass have different corrosion resistances, and (2) a multilayer structural coating with a nanoflake layer (Mg-rich gel layer), a nanowire layer (Ca-rich gel layer), and a porous layer (etched layer) is formed on the air-side of SF glass. Based on the experimental results, we have proposed an aqueous corrosion mechanism of SF glass. The insights of the hydrothermal corrosion behaviors and mechanism provide helpful guidelines to glass surface structural control and functionalization.     

Correlation between instrumental texture and colour quality attributes with sensory analysis of selected cheeses as affected by fat contents

This study evaluated several physical and sensory parameters of different types of cheese available in the Polish market. The measurements of textural properties were conducted in an Instron universal testing machine, while the colour properties of cheeses were measured using a Minolta chromameter. The chemical composition was determined by means of the near‐infrared spectroscopy (NIRs). Moreover, a trained sensory panel was invited to assess the cheese texture‐related properties. Generally, cheeses with reduced fat content were characterised by higher hardness, adhesiveness, cohesiveness and elasticity. Texture‐related parameters of cheese with canola oil were comparable to that of most of full‐fat cheeses. The correlation analysis between physical and sensory attributes related to cheese textural properties indicated the potential applications of TPA, shear and penetration tests (r = 0.766, r = 0.75 and r = 0.765, respectively) for the evaluation of sensory properties related to the hardness. Meanwhile, the elasticity of cheese obtained from sensory evaluation was strongly correlated with the elasticity determined from the shear test (r = 0.722) and moderately correlated with the elasticity from penetration test (r = 0.588), indicating a need to refine the method of penetration test. In addition, cheeses exhibited higher meltability during convection heating at 230 °C than microwave heating. The values of meltability for cheese with reduced fat content were lower than those of full‐fat cheese.