基于深度学习的作物病虫害可视化知识图谱构建

针对作物病虫害领域存在实体关系交叉关联、多源异构数据聚合能力差、知识共享困难等问题，利用知识图谱以结构化的形式描述实体间复杂关系的优势，该研究提出了一种基于深度学习的作物病虫害知识图谱构建方法。该方法在领域本体的基础上，以一种与领域语料相适应的新标注模式实现实体和关系的联合抽取。将实体和关系抽取任务转化为序列标注问题，对实体和关系进行同步标注，有效提高标注效率；为了解决重叠关系抽取问题，直接对三元组建模而不是分别对实体和关系建模，通过标签匹配和映射即可获得三元组数据。利用来自转换器的双向编码器表征量（Bidirectional Encoder Representations from Transformers，BERT）-双向长短期记忆网络（Bi-directional Long-Short Term Memory，BiLSTM）+条件随机场（Conditional Random Field，CRF）端到端模型进行试验，结果表明效果优于基于普通标注方式的流水线方法和联合学习方法中的卷积神经网络（Convolutional Neural Networks，CNN）+BiLSTM+CRF、BiLSTM+CRF等经典模型，F1得分为91.34%。最后，将抽取到的知识存储到Neo4j图数据库中，直观地反映知识图谱的内部结构，实现知识可视化和知识推理。该研究构建的知识图谱可为作物病虫害智能问答系统、推荐系统、智能搜索等下游应用提供高质量的知识库基础。     

Crossing the thresholds: human ecology and historical patterns of landscape degradation

In discussions of landscape sensitivity, human activities have generally been regarded as external forces contributing to landscape change, with a focus on the impacts of cultivation methods, fertiliser practices, grazing pressures and atmospheric pollution. However, there has been comparatively little study undertaken that integrates physical and social systems in a historic context to explain the basis of human activity in sensitive landscapes. Where such attempts have been made, the manner of common land management has figured prominently, with ‘tragedy of the commons’ concepts used to explain land degradation and to provide a foundation for policy response. This has also been the case in Southern Iceland and in this paper we assess the extent to which common land domestic grazing pressures were the primary external force causing soil erosion and land degradation during the period of occupation from ca. 874 AD. We first provide field observation of soil erosion, temporally defined by tephrochronology, to highlight the extent of land degradation during this period. The ‘tragedy of the commons’ explanation of degradation is then assessed by evaluating historic documentary sources, and by environmental reconstruction and modeling of historic grazing pressures. These analyses indicate that regulatory mechanisms were in place to prevent overgrazing from at least the 1200s AD and suggest that there was sufficient biomass to support the numbers of domestic livestock indicated from historic sources. We suggest that failure to remove domestic livestock before the end of the growing season and an absence of shepherding were more likely to contribute to land degradation than absolute numbers. Lack of appropriate regulation of domestic livestock on common grazing areas can be attributed to limited cultural knowledge of changing and rapidly fluctuating environmental conditions.     

Traditional soil and land appraisal on fadama lands in northeast Nigeria

Abstract. As part of a study of recession farming on the 'fadama' lands on a segment of the Komadugu-Yobe floodplain, a survey was carried out to investigate the local farmers' perception of soil types and management practices. The farmers are clearly aware of the differences in soil type on the fadama and they possess unique skills in managing their farm lands. The farmers classify fadama soils for recession farming by assessing soil texture and soil drainage conditions by feel and observation. Integrating such local knowledge into soil surveys will lead to better practical definition of mapping units and give soil names that have more meaning for the farmers.     

农村中学图书馆路在何方?

本文从农村文化素质来看乡镇中学图书馆建设的重要性，探索乡镇中学图书馆的发展之路，并提出了乡镇中学图书馆建设几点建议。     

试论面向知识经济的高校内部审计

知识经济意味着一个双向的社会运动:知识的经济化和经济的知识化。知识经济时代的来临，提高了高校在社会中的地位。知识经济更需要审计监督。知识经济的信息化、数字化、虚拟化和网络化要求使得高校审计工作的目标、方式和内容产生重大变化。高校内部审计需要创新。     

开化县优良乡土树种的选择

本文结合开化县森林植被调查,对开化县的优良乡土树种进行调查选择.采用重要值和频度值指标作为主要依据,结合树种特性分析,选出用材林类造林树种29个,荒山绿化类树种12个,薪炭林造林树种25个,最后就发掘和利用开化县优良乡土树种资源,提出了意见和建议。     

知识互动中的知识交换行为进化博弈分析

利用博弈理论,研究动态知识交互系统中知识个体的知识交换行为特征。引入动态进化博弈链结构，建立知识团队中持续的动态知识交互行为的进化博弈分析模型，在此基础上讨论其可能的演变趋势。最后通过仿真数据分析识别了该行为进化稳定策略（ESS），ESS稳定性特征有利于在知识管理过程中激励员工进行知识交互、把撮知识共享活动中可能的演进规则。     

基于粗糙集理论的立地类型分类规则探讨

利用粗糙集理论（（Roughset））中的知识表达系统和约简算法。在研究区内选取与立地因子关系最密切的树高因子为决策属性，以海拔、母岩、坡度、坡位、土类和土层厚度作为条件属性；决策属性动态聚类和各条件属性值域离散化后，建立立地类型分类决策表，计算决策属性的依赖度及其相对约简，得到马尾松林分立地类型的分类因子为土层厚度、坡位和母岩，并且其重要性分别为0．6923、0．5385和0．3077．     

知识经济时代与办刊——对今后办刊的几点思考

作者简要介绍了“知识经济时代”的由来和特征,指出新时代机遇与挑战的紧迫性,并从农业科技期刊的办刊角度提出具体建议与作法。