基于双向分层语义模型的多源新闻评论情绪预测

随着在线新闻服务的迅猛发展,用户在阅读新闻后可以非常方便地表达自己的主观情绪,有效分析和预测用户的情绪有助于新闻服务提供商为新闻用户提供更好的服务.情绪标注研究已经取得了很多优秀的成果,但仍然存在着一些问题：1)传统的方法将整个文档看作单词流或词袋,不能对句子间的逻辑关系进行建模,在文档中的句子间包含逻辑关系时,这些方法无法适当地表达文档的语义;2)这些方法只用了文档本身的语义,忽略了与该文档相关的其他信息源中信息,而这些信息源对该文档的语义表达也有一定的影响.为了解决这些问题,提出了一种基于多信息源的在线新闻评论双向分层语义表示模型,称为双向分层语义神经网络(bi-directional hierarchical semantic neural network, Bi-HSNN),该模型既捕获句子中词语所表达的情感,又自底向上地学习文档中句子间的逻辑关系,并利用评论、新闻和用户投票等多种信息源对在线新闻评论的情绪进行标注.在真实数据集上的一系列实验,验证了该模型的有效性.     

HSMA：面向物联网异构数据的模式分层匹配算法

随着物联网技术的快速发展,万物互联已经成为必然趋势.物联网技术涉及智能家居、智慧交通等多个领域,使得人们能够随时随地地连接任何人或者设备.然而目前大多数物联网异构数据是孤立的存储,阻碍了万物互联的步伐.数据模式匹配被广泛应用于数据互联并且能够较好地解决以上问题.由于海量物联网数据的异构性以及高增长性,目前的模式匹配方式人工参与度高,与实际应用的契合度低,难以解决物联网新环境下的模式自动匹配问题.提出了一种面向物联网异构数据的模式分层匹配算法,构建3层映射匹配：特征分类匹配、关系特征聚类匹配和混合元素匹配.该算法通过逐层匹配,不断缩小匹配空间,从而提高了匹配质量,减少了元素间匹配次数和人工参与度,较好地实现了自动模式匹配.运用大量数据样本来检验算法的效率和性能,结果证明该算法可行有效.     

深度学习检索框架的前沿探索

经过几十年的发展,信息检索技术获得了长足的进步和广泛的应用,但当前主流的搜索引擎系统距离真正智能的信息获取系统仍然有较大差距.智能信息获取系统能够对网络大数据的内容进行获取、阅读和理解,对关键语义信息实现存储和检索,并能够依据用户的信息需求进行推理、决策和信息生成.实现这样的系统,迫切需要在检索架构和检索模型上形成根本性的改变和理论突破.近年来,围绕智能信息获取的需求,利用深度学习检索框架展开了系统性研究,在数据表征、数据索引以及检索算法等方向上形成了一系列原创成果,在探索全新的深度学习检索架构上不断迈进.     

碱处理对聚偏二氟乙烯结构与介电性能的影响

研究了碱液浸泡处理对聚偏二氟乙烯(PVDF)结构和介电性能的影响。随着浸泡时间的延长,PVDF膜逐渐出现发黄发黑现象。根据傅里叶变换红外光谱、拉曼光谱、X射线光电子能谱等测试表明,PVDF分子链在高温碱液中发生了脱除氟化氢反应,生成碳碳双键,其中部分被继续氧化成含氧基团。接触角测试和宽频介电谱测试结果显示,PVDF膜在强碱中浸泡22 h后,表面水接触角降至31°,亲水性增加;而随着分子链上共轭双键的增加,界面极化程度增大,低频处的介电常数和介电损耗均显著提高。     

城市轨道交通光伏发电系统的总体设计方案与并网模式分析

城市轨道交通已逐渐成为人们日常出行的重要交通工具,将光伏发电系统接入城市轨道交通直流牵引供电系统,不仅可以节能降耗,推动能源供应结构的转型,同时还能促进当地太阳能资源的利用和开发。首先概述了城市轨道交通光伏发电系统的总体设计方案,然后对光伏发电系统几种并网模式展开了讨论,最后对其今后的应用前景及发展方向进行了分析,以供该领域相关人员参考。     

A Stress Self-Adaptive Bimetallic Stellar Nanosphere for High-Energy Sodium-Ion Batteries

Bimetallic composites exhibit great potential as anode materials in advanced energy storage systems owing to their inherent tunability, cost-effectiveness, and simultaneous achievement of high specific capacity and low reaction potential. However, simple biphase mixing often fails to achieve satisfactory performance. Herein, an innovative stress self-adaptive bimetallic stellar nanosphere (50–200 nm) wherein bismuth (Bi) is fabricated, as a core, is seamlessly encapsulated by a tin (Sn) sneath (Sn-Bi@C). This well-integrated stellar configuration with bimetallic nature embodies the synergy between Bi and Sn, offering fortified conductivity and heightened sodium ion diffusion kinetics. Moreover, through meticulous utilization of finite element analysis simulations, a homogeneous stress distribution within the Sn-enveloped Bi, efficiently mitigating the structural strain raised from the insertion of Na⁺ ions, is uncovered. The corresponding electrode demonstrates remarkable cyclic stability, as it exhibits no capacity decay after 100 cycles at 0.1 A g⁻¹. Furthermore, it achieves an impressive 86.9% capacity retention even after an extensive 2000 cycles. When employed in a Na₃V₂(PO₄)₃ ‖ Sn-Bi@C full cell configuration, it demonstrates exceptional capacity retention of 97.06% after 300 cycles at 1 A g⁻¹, along with a high energy density of 251.2 W h kg⁻¹.     

Inch-Size Achiral Perovskite Single Crystals for Distinguishing Circularly Polarized Light with A Large Asymmetry Factor

Circularly polarized light (CPL) detection has sparked the booming interest for versatile chiral-optoelectronic and spintronic devices. For practical device applications, a large asymmetry factor is required to distinguish the handedness of incent light. However, for most chiral CPL active candidates, the difference in optical density or absorption anisotropy is quite subtle under CPL illumination, resulting in a relatively low asymmetry factor (usually ≤ 0.2). Up to date, it remains extremely challenging to achieve CPL response in the achiral crystals. Inch-size achiral perovskite-type crystals of (IBA)₂(EA)₂Pb₃I₁₀ (IBA = 4-isopropylbenzylammonium, EA = ethylammonium), which holds a promise to distinguish CPL with a large asymmetry factor up to ≈0.56 at 405 nm is presented here. This figure-of-merit is comparable or even superior than most of those chiral counterparts. Further First-principles calculations disclose such CPL-active behaviors of (IBA)₂(EA)₂Pb₃I₁₀ are associated with the splitting of its band spin-degeneracy, induced by the breaking of inversion symmetry and spin-orbital coupling (SOC) effects. As far as it is known, the structural origin of CPL-activities of (IBA)₂(EA)₂Pb₃I₁₀ is completely distinctive from the conventional chiral family, which paves a promising pathway to explore new candidates toward high-performance CPL detection.     

Direct Photolithography Patterning of Quantum Dot-Polymer

For the new display technology based on quantum dots (QDs), achieving high-precision red, green, and blue pixel arrays has always been a research focus in the pursuit of high-quality and vivid image displays. However, problems such as material stability and process environment make it difficult to guarantee the quality of high-precision patterns. The new optical patterning technology represented by direct photolithography is considered a highly promising method for achieving ultrafine patterns at the submicron level. This process prepares patterned quantum dot-polymer films through light-induced chemical changes. This paper reviews the progress of direct photolithography research focused on QD-polymer materials and presents recent advances in such processes for monochromatic/multicolor light patterning. The article classifies QD-polymers into three categories by combining QDs with polymers in different ways, including polymer-coated QDs, polymers as QD ligands, and polymers as photocrosslinkers for QDs. Their synthesis schemes, functional features, and challenges are also presented. In addition, a scheme to remove the photomask during direct lithography using lasers and light field modulation is also proposed. It aims to provide readers and researchers with some generalized research information and improvement ideas. This can further advance the development of direct photolithography for QD-polymers.     

Feedforward Photoadaptive Organic Neuromorphic Transistor with Mixed-Weight Plasticity for Augmenting Perception

Organic photoelectric neuromorphic devices that mimic the brain are widely explored for advanced perceptual computing. However, current individual neuromorphic synaptic devices mainly focus on utilizing linear models to process optoelectronic signals, which means that there is a lack of effective response to nonlinear structural information from the real world, severely limiting the computational efficiency and adaptability of networks to static and dynamic information. Here, a feedforward photoadaptive organic neuromorphic transistor with mixed-weight plasticity is reported. By introducing the potential of the space charge to couple gate potential, photoexcitation, and photoinhibition occur successively in the channel under the interference of constant light intensity, which enables the device to transform from a linear model to a nonlinear model. As a result, the device exhibits a dynamic range of over 100 dB, exceeding the currently reported similar neuromorphic synaptic devices. Further, the device achieves adaptive tone mapping within 5 s for static information and achieves over 90% robustness recognition accuracy for dynamic information. Therefore, this work provides a new strategy for developing advanced neuromorphic devices and has great potential in the fields of intelligent driving and brain-like computing.     

Effect of Overexpression of γ-Tocopherol Methyltransferase on α-Tocopherol and Fatty Acid Accumulation and Tolerance to Salt Stress during Seed Germination in Brassica napus L.

Rapeseed (Brassica napus L.) is an important oil crop and a major source of tocopherols, also known as vitamin E, in human nutrition. Enhancing the quality and composition of fatty acids (FAs) and tocopherols in seeds has long been a target for rapeseed breeding. The gene γ-Tocopherol methyltransferase (γ-TMT) encodes an enzyme catalysing the conversion of γ-tocopherol to α-tocopherol, which has the highest biological activity. However, the genetic basis of γ-TMT in B. napus seeds remains unclear. In the present study, BnaC02.TMT.a, one paralogue of Brassica napus γ-TMT, was isolated from the B. napus cultivar “Zhongshuang11” by nested PCR, and two homozygous transgenic overexpression lines were further characterised. Our results demonstrated that the overexpression of BnaC02.TMT.a mediated an increase in the α- and total tocopherol content in transgenic B. napus seeds. Interestingly, the FA composition was also altered in the transgenic plants; a reduction in the levels of oleic acid and an increase in the levels of linoleic acid and linolenic acid were observed. Consistently, BnaC02.TMT.a promoted the expression of BnFAD2 and BnFAD3, which are involved in the biosynthesis of polyunsaturated fatty acids during seed development. In addition, BnaC02.TMT.a enhanced the tolerance to salt stress by scavenging reactive oxygen species (ROS) during seed germination in B. napus. Our results suggest that BnaC02.TMT.a could affect the tocopherol content and FA composition and play a positive role in regulating the rapeseed response to salt stress by modulating the ROS scavenging system. This study broadens our understanding of the function of the Bnγ-TMT gene and provides a novel strategy for genetic engineering in rapeseed breeding.