基于部首嵌入和注意力机制的病虫害命名实体识别

为了解决农业病虫害命名实体识别过程中存在的内在语义信息缺失、局部上下文特征易被忽略和捕获长距离依赖能力不足等问题，以农业病虫害文本为研究对象，提出一种基于部首嵌入和注意力机制的农业病虫害命名实体识别模型(Chinese agricultural diseases and pests named entity recognition with joint radical embedding and self attention, RS-ADP)。首先，该模型将部首嵌入集成到字符嵌入中作为输入，用以丰富语义信息。其中，针对部首嵌入设计了3种特征提取策略，即卷积神经网络(Convolutional neural network, CNN)、双向长短时记忆网络(Bidirectional long short term memory network, BiLSTM) 和CNN-BiLSTM；其次，采用多层不同窗口尺寸的CNNs层提取不同尺度的局部上下文信息；然后，在BiLSTM提取全局序列特征的基础上，采用自注意力机制进一步增强模型提取更长距离依赖的能力；最后，采用条件随机场(Conditional random field, CRF)联合识别实体边界和划分实体类别。在包含11个类别和24715条标注样本的农业病虫害自制语料上进行了实验。结果表明，本文模型RS-ADP在该数据集上精确率、召回率和F1值分别为94.16%、94.47%和94.32%；在具体实体类别上，RS-ADP在作物、病害、虫害等易识别实体上F1值高达95.81%、97.76%和97.23%。同时，RS-ADP在草害、病原等难以识别实体上F1值仍保持86%以上。实验结果表明，本文所提模型能够有效识别农业病虫害命名实体，其识别精度优于其他模型，且具有一定的泛化性。     

高层建筑和低层裙房基础设计方案选择

针对高层建筑和低层裙房基础设计中常遇到的方案选择、高层主楼与低层裙房基础沉降差异处理等问题，提出了设计要求和实施的技术措施。     

梨三倍体新品种龙园洋红的选育

龙园洋红梨是梨三倍体新品种,亲本为56-520×乔玛,抗寒性强,短枝型.果实不规则长椭圆形,果形指数1.13,平均单果重185.5 g,每个果实平均2.5粒种子,种子小、不饱满,多畸形.果皮浅黄色,阳面有红晕,果肉乳白色,肉质细软多汁,风味甜,有香气,可溶性固形物16.05%.果实在哈尔滨地区9月中旬成熟.     

装修与电气

本文着重地论述了由装修而引发的电气安全技术问题已到了不容忽视的程度。就此问题作 者提出了自己的观点及解决办法。     

两端不等弯矩作用下薄壁梁柱的弯扭屈曲

本文采用Ｒａｙｌｅｉｇｈ－Ｒｉｔｚ法求解两端不等弯矩作用下单轴对称开口薄壁截面梁柱弯扭屈曲临界荷载，分析中考虑了三种常见的边界条件以及屈曲前变形的影响，将本文弯扭屈由临界荷载理论值与试验值比较，吻合较好。     

变截面构件的剪应力计算

变截面构件的弯曲变形是实际构件常见变形之一，现行钢筋混凝土设计规范提供的和实际设计计算应用的剪应力计算公式未考虑截面变化的影响，有较大的误差。本文用材料力学方法推出考虑变截面影响的剪应力计算公式，以便在实际设计计算应用。     

开口薄壁结构弯扭屈曲能量原理的有限变形理论分析

The formulas of the total potential energy of arbitrary thin- walled open cross section member counting in the geometric non-linear factors is deduced. The deductions are based on the most general finite deformation theory and the two assumptions that"the     

新亚欧大陆桥新疆段受损区段的灰色预测

在对新亚欧大陆桥新疆段受损区段、受损长度及受损频率和受损密度详尽分析的基础上，揭示了陆桥新疆段受损区段的时空分布规律。并依据灾害的时序分布特征，运用灰色原理和方法，根据受损区段数确立阈值和相应的年份序列，分别建立GM（1，1）模型，求出年份序列的序号响应式，并对实际值和理论预测结果进行了拟合检验，模型群的原点误差范围为0.198-1.314之间，平均误差为0.823-1.917，并同时预测了陆桥新疆段2001－2005年间可能的受损区段数。     

Gas and short‐chain fatty acid production from feeds commonly fed to red deer (Cervus elaphus L.) and incubated with rumen inoculum from red deer and sheep

Efficient red deer supplementary feeding depends on estimations of the nutritive value of offered feeds, frequently estimated with the use of equations derived from domestic ruminants. The aim of this study was to compare the 24‐hour in vitro true dry matter degradability (ivTD₂₄), in vitro gas production (GP) kinetic parameters, GP in 24 hr of incubation (GAS₂₄) and short‐chain fatty acid (SCFA) and microbial biomass (MBS) produced after 24‐hour incubation of feeds in inoculum prepared from sheep and red deer rumen fluid. Eleven feeds, frequently consumed by red deer in Slovenia, which occur either naturally (two fresh grasses, chestnut fruits and common and sessile oak acorns) or are fed as winter supplemental feeds (two grass hays, two grass silages, apple pomace, fresh sugar beetroot), were investigated. The in vitro GP kinetic parameters, GAS₂₄ and ivTD₂₄, did not differ between animal species. Amounts of SCFAs were greater (p < 0.05) when feeds were incubated in sheep inoculum, while molar proportions of acetic and propionic acids did not differ. Molar proportions of butyric acid produced during incubation of high fibre feeds did not differ between animal species, but were higher (p < 0.05) when feeds high in starch or sugar were incubated in red deer inoculum. Greater production of SCFA by sheep rumen microbes suggests better coverage of host animal with energy precursors, while greater production of MBS by red deer rumen microbes suggests better coverage of host animal with protein. Results also suggest that rumens of sheep and red deer are inhabited by different microbial communities, which did not affect the extent of in vitro GP and degradation of feeds used in the present experiment. However, the possibility exists that the divergent nutrient use could be a consequence of different priming by different feeds of the donor animal diets.