基于BERT-BILSTM-CRF模型的电力行业事故文本智能分析*

为解决电力行业事故报告文本较长、语义复杂,难以进行有效文本识别问题,提出1种以BERT作为底层的预训练模型,并设计1种双重注意力机制编码器,结合BILSTM-CRF深度挖掘事故文本语义特征,从而实现文本智能分析。首先构建电力词典,通过对BERT预训练,进行BIO标注,然后引入BILSTM-CRF模型实现对文本标签智能分类,最后将该模型与现行其他4种深度学习模型进行对比。研究结果表明:该模型智能识别精确率、召回率及F1值(查准率)均达到约97%,较其他4种模型中效果最好的模型分别提高0.02,0.03,0.02。研究结果可为电力行业事故报告文本分析提供1种新思路。     

基于缓冲区重采样的LSTM林火预测模型*

为提高林火风险预测精度,挖掘地图上隐含的空间信息、时间序列上隐含的长期趋势和循环波动,提出1种基于缓冲区重采样的长短期记忆（LSTM）林火预测模型,选取15个与林火相关的影响因素,以方差膨胀因子为评价指标对其进行多重共线性检验,方差膨胀因子大于10的因素具有共线性,并采用信息增益率验证筛选结果的合理性。考虑到火灾的空间聚集特性,采用缓冲区分析与过采样相结合方法减少样本不均衡现象的影响,最终得到176 732条样本。对12个影响因素和研究时间段的火点建立LSTM预测模型,对森林火灾发生风险进行预测。研究结果表明:基于缓冲区重采样的LSTM林火预测模型有效考虑时空上隐含的信息,预测模型准确率为87.06%,特异性为97.99%,敏感度为76.12%,阳性预测率为97.43%,阴性预测率为80.41%,ROC曲线与AUC值均优于随机森林（RF）和支持向量机（SVM）这2种基准算法。维尔克松秩和检验发现,本文提出的模型与基准算法结果具有显著性差异。研究结果可为提高林火风险预测精度提供参考。     

“Science is Everywhere,but No One Knows It”: Assessing the Cultural Distance to Science of Rural South African Publics

ABSTRACT

In science communication, a prerequisite for reaching different parts of a society is to find out how these publics experience and interpret science. Since rural South African publics are perceived to exhibit a large cultural distance to science, the present exploratory study aimed to know in more detail how rural South Africans perceive and understand science in their local and social contexts. Theoretical notions on cultural distance and the methodological approach of segmentation studies were considered. Semi-structured interviews with rural South Africans were carried out in four towns (n?=?52) that differ with respect to having a large scientific installation in their vicinity, or not. Sensitively comparing local and social contexts helped identifying three different publics who differed regarding their perceptions of science; however, the large scientific installations only made a difference in perceptions for those publics who were generally less exposed to and less knowledgeable about science.     

泉州文化旅游资源特色及其开发利用

在分析“世界宗教博物馆”泉州的海外交通、宗教、建筑和民俗文化旅游资源特色及其优劣势的基础上，就其开发利用过程中通过整合资源和拓展内涵来提炼和突出文化主题，增强文化旅游的参与性，加强从业人员教育和重视品牌塑造等方面提出建议。     

邻苯二甲酸二丁酯影响阿尔茨海默病大鼠学习记忆能力的分子机制研究

为探究邻苯二甲酸二丁酯(DBP)影响阿尔茨海默病(AD)大鼠学习记忆能力的分子机制,将56只SPF级5～6周龄雌性Wistar大鼠随机平均分为7组:阴性对照组、0.25 mg·kg-1·d-1 DBP 组、2.5 mg·kg-1·d-1 DBP 组、25 mg·kg-1·d-1 DBP 组、250 mg·kg-1·d-...     

PPARγ/UCP2在甲醛致学习记忆障碍中的作用

为探究过氧化物酶体增殖物激活受体 γ/解偶联蛋白2(PPARγ/UCP2)在甲醛(FA)诱导的学习记忆障碍中的作用,本文将C57BL/6小鼠随机分为:对照组、T0070907组(抑制剂组)、3mg/m3FA组、3mg/m3FA+T0070907组,进行连续21d的实验暴露,在第22d取脑组织测定脑组织脏体比并匀浆,检测...     

中央苏区红色旅游资源跨区域整合开发研究：文化生态学视角

红色旅游资源是革命精神的核心载体,其保护和旅游利用面临着破碎化、片段化、静态化等问题,亟需开展整合开发研究。以中央苏区核心区为例,基于文化生态学理论,采用GIS空间分析方法和修正的引力模型,研究红色旅游资源空间格局和文化生态变迁,在此基础上提出整合开发的策略,结论如下：（1）红色旅游资源点空间上呈现局部聚集特征,赣南苏区形成章贡区、瑞金市、于都县“金三角”分布格局,闽西苏区则形成新罗区、永定区、上杭县、连城县“银菱形”分布格局;（2）闽西和赣南苏区各重大事件红色旅游资源点在空间上8.9~106.6 km的范围内,形成带状或面状的文化生态基底,并呈现一致向服务业主导型转变的发展态势;（3）红色旅游资源网络联系方向与服务业主导型文化生态方向错位发展,这是赣南和闽西红色旅游资源内部网络联系度较强,跨区域网络联系度较弱的重要原因。在此基础上,遵循红色旅游资源文化生态空间一体化、文化生态演变全域化、文化生态发展动态化的原则,提出空间集聚区优先整合开发、高发展指数红色旅游资源跨省域优先整合开发、跨主导类型关键节点优先整合开发三个整合开发对策,以期促进中央苏区红色旅游高质量发展。     

大气污染对石灰质文物的风化影响及防护新材料

综述了大气污染对石灰质文物风化影响的研究进展,涉及大气中气体污染物、大气中酸性降水、大气中的气溶胶。文章介绍了目前开发的防护新材料,主要包括有机氟聚合物材料、纳米材料和仿生材料,并指出开发未来新型防护材料的方向是寻找安全生物钙化菌和开发与石灰质文物相匹配的新型保护材料。     

Why social values cannot be changed for the sake of conservation

The hope for creating widespread change in social values has endured among conservation professionals since early calls by Aldo Leopold for a “land ethic.” However, there has been little serious attention in conservation to the fields of investigation that address values, how they are formed, and how they change. We introduce a social–ecological systems conceptual approach in which values are seen not only as motivational goals people hold but also as ideas that are deeply embedded in society's material culture, collective behaviors, traditions, and institutions. Values define and bind groups, organizations, and societies; serve an adaptive role; and are typically stable across generations. When abrupt value changes occur, they are in response to substantial alterations in the social–ecological context. Such changes build on prior value structures and do not result in complete replacement. Given this understanding of values, we conclude that deliberate efforts to orchestrate value shifts for conservation are unlikely to be effective. Instead, there is an urgent need for research on values with a multilevel and dynamic view that can inform innovative conservation strategies for working within existing value structures. New directions facilitated by a systems approach will enhance understanding of the role values play in shaping conservation challenges and improve management of the human component of conservation.     

生物表面活性剂产生菌犁头霉菌(Absidia orchidis)的筛选及发酵条件优化

采用植物油为唯一碳源,设计选择培养基,从饭店下水道污泥中筛选出生物表面活性剂产生菌.结果分离到12株菌,其中一株能使发酵液的表面张力值从68 mN·m-1降低到34.5 mN·m-1,具有开发潜力,被选出作进一步的研究.该菌株经鉴定为犁头霉菌(Absidia orchidis).通过正交试验对犁头霉菌的培养条件进行优化,其优化培养条件为:植物油3.6 g·L-1,KH2PO412.1 g·L-1,Na2HPO45 g·L-1,(NH4)2SO4 1 g·L-1,NaNO32 g·L-1,酵母浸膏0.1 g·L-1,MgSO4·7H2O 0.15 g·L-1,NaCl 5 g·L-1,CaCl2 0.1 g·L-1,EDTA 1 g·L-1,KI 0.83μg·L-1,H3PO4 0.01μg·L-1,CoCl2·6H2O 0.048μg·L-1,MnSO4·H2O 0.312μg·L-1,Na2 MoO4·2H2O 0.048μg·L-1,ZnCl2 0.287μg·L-1,CuSO4·5H2O 0.125μg·L-1,初始pH值8,接种量6%.发酵70h时可获得生物表面活性剂的最大收获量,此时发酵液中生物表面活性剂的相对浓度达402倍.