ASE-ERNIE:一种基于ERNIE的中文指令动作序列生成方法

中文结构灵活多样,同样的语义有多种表达方式,在中文指令过长或存在多组动作的情况下,家庭服务机器人很难正确理解并做出响应。此外,家庭环境的中文指令解析研究在数据方面仍然稀缺。现有的方法不能准确提取出指令中包含的全部动作序列。由此提出一种基于ERNIE方法的动作序列提取(ASE-ERNIE)模型来提取中文指令中的可执行动作序列,该模型通过动作类型和参数之间的关系提取指令信息,将动作序列的提取工作转化为序列标注任务。在指令中存在多组动作并且存在多个参数的情况下,依然能够提取出全部动作序列。通过实验验证了所提方法的有效性,动作序列提取任务的F 1值达到了80.37%。

基于ERNIE-AA模型的诉求文本分类技术

12345市民服务热线中汇集了大量的市民诉求,对这些诉求文本进行分类能够有效地提高市民服务热线的办事效率。为了快速且精准地对诉求文本进行分类,提出一种融合了人工注意力机制的ERNIE-AA模型。该模型将人工分类时的启发式规则通过人工注意力机制引入ERNIE模型的输入表示,通过预训练模型增加影响分类的关键词的注意力权重。实验结果表明:ERNIE-AA模型在诉求文本数据集上具有良好的表现,分类准确率可达86.71%。

一种基于内容和ERNIE3.0-CapsNet的中文垃圾邮件识别方法

针对目前中文垃圾邮件识别方法中的深度学习检测方法词向量表示不足和特征提取丰富度欠缺的问题,提出融合ERNIE3.0预训练模型的胶囊神经网络改进识别模型——ERNIE3.0-CapsNet.对于中文垃圾邮件内容文本,利用ERNIE3.0生成对于知识具备优异记忆和推理能力且语义丰富的词向量矩阵,再使用胶囊神经网络进行特征提取及分类,对于胶囊神经网络,改进了结构并使用GELU作为其动态路由的激活函数,设计了5组同类模型和4组激活函数的对比实验.在开源的TREC06C中文邮件数据集上,提出的ERNIE3.0-CapsNet模型效果在总体上表现突出,其准确率达到99.45%.实验结果表明,ERNIE3.0-CapsNet优于ERNIE3.0-TextCNN,ERNIE3.0-RNN等方法,证明了该模型在中文垃圾邮件识别效果的有效性和优异性.

基于ERNIE及改进DPCNN的棉花病虫害问句意图识别

针对目前没有公开的棉花病虫害相关问句数据集且问句较短、类型多样等问题,本研究通过查阅文献及咨询相关领域专家,构建了棉花病虫害问句数据集CQCls,定义了78种棉花病虫害实体和9种问句类型;同时提出了一种基于ERNIE预训练模型的棉花病虫害问句意图识别模型,首先通过ERNIE模型将输入问句映射到向量空间,使用融合词位置信息的DPCNN模型进行特征向量的抽取,与基础的DPCNN模型相比,通过融合词位置信息能有效提高模型的表达能力,然后经过Softmax得到最终结果。实验结果表明,本研究提出的意图识别模型相较于其他模型取得了较好的结果,宏平均和加权平均的F1分数值分别为97.45%和97.31%;在文本语料数据内容复杂多样且文本格式不规范的DMSCD数据集上,训练结果中不同类别的F1分数的权重平均也能达到73.42%,进一步证明了该模型的有效性及泛化能力。

基于ERNIE-BiGRU-Attention-CRF的电子病历命名实体识别方法

目的/意义改善中文电子病历命名实体识别模型的性能,更好地开展医疗信息的组织和挖掘。方法/过程构建ERNIE-BiGRU-Attention-CRF中文电子病历命名实体识别模型,首先采用ERNIE1.0预训练模型生成具有语义特征的词向量,然后利用BiGRU捕获全局语义特征与语法结构特征,通过Attention机制进一步增强语义特征的捕获,最后连接CRF解码层输出全局概率最大的标签序列。结果/结论在公开的医疗文本数据集CCKS2017开展对比实验、消融实验,利用生成的模型进行实例分析,取得较好的识别效果。

基于知识图谱联合ERNIE-DPCNN模型的药品不良反应自动关联性评价方法研究

目的 针对当前药品不良反应关联性评价存在的效率较低和主观性评估问题,通过建立药品不良反应关联性评价模型,探索药品不良反应自动关联性评价方法。方法 利用文献及互联网来源,对获取的不良反应报告标注数据(7 301条)进行知识抽取,构建药品不良反应知识图谱,建立知识驱动的ERNIE-DPCNN自动关联性评价模型。结果 提出的知识图谱联合ERNIE-DPCNN模型在测试集中的精确度、召回率和F1值分别达到82.18%、81.40%、81.21%,相对于其他基线模型各项评估指标均取得了最高值。结论 知识图谱联合ERNIE-DPCNN模型的方法能提高药品不良反应关联性评价效率,具备较强的准确性,并在一定程度上减少主观性评估误差,对基于人工智能的自动化评价有一定参考价值。

融合ERNIE和双向注意力流的医生推荐算法

医生推荐是在线问诊平台的研究热点之一,为了向患者提供更加精准、个性化的医生推荐,提出一种融合ERNIE和双向注意力流的医生推荐算法。首先,利用ERNIE预训练模型对患者疾病描述进行词向量编码,通过弱监督的方式自定义标签生成医生擅长文本训练数据,再将医生职称信息与擅长文本进行拼接输入到ERNIE模型中;其次,采用双向注意力流(bi-directional attention flow)捕捉医生文本中的关键信息,学习患者疾病描述文本和医生文本的关系,获得融合患者信息的医生综合特征向量;最后,将医生综合特征和患者特征进行相似度计算产生推荐结果。利用来自“好大夫在线”的真实数据进行了不同算法实验对比,结果表明所提算法在推荐性能上有明显提升。

Differential response of radial growth and δ^(13)C in Qinghai spruce(Picea crassifolia) to climate change on the southern and northern slopes of the Qilian Mountains in Northwest China

Tree radial growth can have significantly differ-ent responses to climate change depending on the environ-ment.To elucidate the effects of climate on radial growth and stable carbon isotope(δ^(13)C)fractionation of Qing-hai spruce(Picea crassifolia),a widely distributed native conifer in northwestern China in different environments,we developed chronologies for tree-ring widths and δ^(13)C in trees on the southern and northern slopes of the Qilian Mountains,and analysed the relationship between these tree-ring variables and major climatic factors.Tree-ring widths were strongly influenced by climatic factors early in the growing season,and the radial growth in trees on the northern slopes was more sensitive to climate than in trees on the southern.Tree-ring δ^(13)C was more sensitive to climate than radial growth.δ^(13)C fractionation was mainly influenced by summer temperature and precipitation early in the growing season.Stomatal conductance more strongly limited stable carbon isotope fractionation in tree rings than photosynthetic rate did.The response between tree rings and climate in mountains gradually weakened as climate warmed.Changes in radial growth and stable carbon isotope fractionation of P.crassifolia in response to climate in the Qilian Mountains may be further complicated by continued climate change.

基于ERNIE-BiGRU-CRF模型的煤矿安全隐患命名实体智能识别研究

为充分挖掘煤矿安全隐患文本关键知识,帮助煤矿企业安全管理人员更好的开展隐患排查治理工作,提出一种基于预训练语言模型的命名实体识别方法。首先定义煤矿安全隐患实体类别,并采用BIO标注策略构建了7个实体类别和15个实体标签;然后将收集到的煤矿隐患排查数据进行预处理,由煤矿安全领域专家人工标注相关实体,得到1500条煤矿安全隐患命名实体标准数据集;最后采用ERNIE预训练模型对煤矿安全隐患文本词向量进行表征、同时利用BiGRU结构进行上下文语义特征提取以及CRF模型进行实体标签解码,完成煤矿安全隐患命名实体识别研究。实验结果表明:ERNIE-BiGRU-CRF模型在序列标注任务上的精确率、召回率和F1值分别为56.69%、69.23%和62.34%,较于BiLSTM-CRF基线模型分别提高了6.85%、13.74%和9.83%,并且实体抽取结果与实际标注结果相差不大。另外,消融实验也验证了BiGRU层能够更好的捕捉煤矿安全隐患文本上下文语义依赖关系以及CRF层能够进一步优化标签序列的有效性。

Present-day movement characteristics of the Qinghai Nanshan fault and its surrounding area from GPS observation

The Qinghai Nanshan fault is a larger fault in the Northeastern Xizang Plateau.In previous studies,its movement characteristics are mainly investigated with geological and seismic observations,and the tectonic transformation role of the fault on its east is not yet clear.This study uses data fusion to obtain denser GPS observations near the Qinghai Nanshan fault.Based on tectonic characteristics,we establish a block model to investigate the fault slip rate,locking degree,and slip deficit.The results show that the Qinghai Nanshan fault slip rate is characterized by sinistral and convergent movement.Both the sinistral and convergent rates display a decreasing trend from west to east.The locking degree and slip deficit are higher in the western segment(with an average of about 0.74 and 1.1 mm/a)and lower in the eastern segment.Then,we construct a strain rate field using GPS observations to analyze the regional strain characteristics.The results indicate that along the fault,the western segment shows a larger shear strain rate and negative dilation rate.Regional earthquake records show that the frequency of earthquakes is lower near the fault.The joint results suggest that the western segment may have a higher earthquake risk.In addition,the insignificant fault slip rate in the eastern segment may indicate that it does not participate in the tectonic transformation among the Riyueshan,Lajishan,and West Qinling faults.

Quantification of the concrete freeze–thaw environment across the Qinghai–Tibet Plateau based on machine learning algorithms

The reasonable quantification of the concrete freezing environment on the Qinghai–Tibet Plateau(QTP) is the primary issue in frost resistant concrete design, which is one of the challenges that the QTP engineering managers should take into account. In this paper, we propose a more realistic method to calculate the number of concrete freeze–thaw cycles(NFTCs) on the QTP. The calculated results show that the NFTCs increase as the altitude of the meteorological station increases with the average NFTCs being 208.7. Four machine learning methods, i.e., the random forest(RF) model, generalized boosting method(GBM), generalized linear model(GLM), and generalized additive model(GAM), are used to fit the NFTCs. The root mean square error(RMSE) values of the RF, GBM, GLM, and GAM are 32.3, 4.3, 247.9, and 161.3, respectively. The R^(2) values of the RF, GBM, GLM, and GAM are 0.93, 0.99, 0.48, and 0.66, respectively. The GBM method performs the best compared to the other three methods, which was shown by the results of RMSE and R^(2) values. The quantitative results from the GBM method indicate that the lowest, medium, and highest NFTC values are distributed in the northern, central, and southern parts of the QTP, respectively. The annual NFTCs in the QTP region are mainly concentrated at 160 and above, and the average NFTCs is 200 across the QTP. Our results can provide scientific guidance and a theoretical basis for the freezing resistance design of concrete in various projects on the QTP.

Origin of the Dashuigou independent tellurium deposit at Qinghai–Xizang Plateau: constraints from the light stable isotopes C, O, and H

By studying the light isotopic compositions of carbon,oxygen,and hydrogen,combined with previous research results on the ore-forming source of the deposit,the authors try to uncover its metallogenic origin.The δ^(18)O and δ^(13)C isotope signatures of dolomite samples vary between 10.2 and 13.0‰,and between−7.2 and−5.2‰,respectively,implying that the carbon derives from the upper mantle.δD and δ^(18) O of quartz,biotite,and muscovite from diff erent ore veins of the deposit vary between−82 and−59‰,and between 11.6 and 12.4‰,respectively,implying that the metallogenic solutions are mainly magmatic.According to the relevant research results of many isotope geologists,the fractionation degree of hydrogen isotopes increases as the depth to the Earth’s core increases,and the more diff erentiated the hydrogen isotopes are,the lower their values will be.In other words,mantle-derived solutions can have extremely low hydrogen isotope values.This means that the δD‰ value−134 of the pyrrhotite sample numbered SD-34 in this article may indicate mantle-derived oreforming fl uid of the deposit.The formation of the Dashuigou tellurium deposit occurred between 91.71 and 80.19 Ma.

Fluid inclusion and H-O-S-Pb isotope systematics of the Chayong Cu-polymetallic deposit,Sanjiang Metallogenic Belt,Qinghai Province,China

The Chayong Cu-polymetallic deposit is a recently discovered Cu-polymetallic deposit hosted in the Sanjiang Metallogenic Belt within the Tibetan Plateau of China to the northeast of the North Qiangtang terrane.The ore body occurs in siltstone and is controlled by a northwest-trending fault structure.According to the associations,assemblages,and cutting relationships between ore veins,the hydrothermal mineralization period can be divided into three mineralization stages:(1)a molybdenite mineralization stage,(2)a Cu-polymetallic sulfide stage,and(3)a quartzcarbonate stage.Two types of fluid inclusions(FIs),namely,liquid and vapor-rich inclusions,are present in quartz as so ciated with sulfide minerals.Early-stage FIs are both iquid and vapor-rich,homogenized at temperatures ranging from 364.1 to 384.2℃,and have salinities ranging from0.70%to 9.60%NaCl equivalent(eqv).The middle-stage FIs are also both liquid-and vapor-rich,homogenized at temperatures ranging from 272.4 to 355.6℃,and have salinities ranging from 0.53%-17.10%NaCl eqv.The late-stage FIs are liquid,homogenized at temperatures ranging from 209.4to 255.3℃,and have salinities ranging from 0.35%-6.87%NaCl eqv.The samples from the deposit haveδ^(34)S values of-21.8‰to-19.2‰and-5.5‰to-6.0‰,suggesting that sulfur was derived from the host sediments and magmatic fluids,respectively.The metallic minerals within the deposit have^(206)Pb/^(204)Pb,^(207)Pb/^(204)Pb,and^(208)Pb/^(204)Pb values of 18.439-18.458,15.656-15.679,and 38.772-38.863,respectively,suggesting that the metals were derived from the upper crust and orogenic belts.The samples from the deposit haveδ^(18)O_(W)values of 2.99‰-7.99‰andδD_(W) values ranging from-84.4‰to-73.9‰,indicating that the pre-forming fluids were magmatic and mixed with minor amounts of meteoric water.The ore-forming fluid of the Chayong copper polymetallic deposit was a high-temperature,medium-to low-salinity H_(2)O-NaCl-CH_(4)-N_(2)±CO_(2)fluid system.The early high-temperature magmatic fluid,due to boiling,decreased in temperature,and via the mixing of meteoric water,gradually evolved towards the later-stage medium-to low-temperature and low-salinity fluid,causing nolybdenite mineralization and forming copper polymetallic sulfide veins and quartz carbonate veins.

Spatial-temporal heterogeneity of ecological quality changes across the Qinghai-Tibet Plateau under the influence of climate factors and human activities

Over the last few decades,the ecological quality of the Qinghai–Tibet Plateau(QTP)has significantly changed due to climate warming,humidification,and increasing human activities.Thus,evaluating this region's ecological quality and dominant factors is crucial for sustainable development.In this study,the changes in the ecological quality on the QTP from 2000 to 2020 were evaluated based on aggregated indices and Sen–MK trend analyses,and the dominant factors affecting the ecological quality of the QTP were quantitatively analyzed using decision tree classification.The results revealed that(1)the ecological quality of the QTP exhibited an overall high trend in the east and a low pattern in the west;(2)the ecological quality of the QTP significantly increased from 2000 to 2020,and human activities were the dominant factors causing this change;and(3)the changes in the ecological quality and dominant factors exhibited obvious spatiotemporal heterogeneity.The area with an improved ecological quality occurred mainly in the northern QTP region.It was governed by human activities and precipitation.In contrast,the area with a deteriorated ecological quality occurred largely in the southern QTP region and was dominated by human activities and temperature.The 2000–2010 period was the most significant period of heterogeneity regarding of ecological quality and its driving factors.(4)The change in the ecological quality was mainly affected by the synergistic relationship between human activities and climate change in this region,which encompassed multiple dominant factors.This study provides important information on the spatiotemporal heterogeneity of ecological quality change and its dominant factors on the QTP and offers systematic guidance for the planning and implementation of ecological protection projects.

基于ERNIE-BiAGRU-CapsNet的中文评教文本情感分析

为提高中文评教文本情感分析的准确率并挖掘文本中深层次语义情感关联,文章提出了一种基于ERNIE-BiAGRU-CapsNet的中文评教文本情感分析模型。其中,ERNIE通过词嵌入层获取文本中的隐含关系及深层语义动态词向量;BiAGRU在BiGRU更新门加入了注意力机制,以深入挖掘文本上下文语义表征信息;针对提取的上下文语义特征,CapsNet旨在进一步挖掘其局部语义特征。通过在实际中文评教文本数据中进行实验验证,模型分类输出的精确率高达96.77%,其分类性能较同类深度学习模型更优。

基于ERNIE-BiLSTM-DPCNN的微博长文本谣言检测

为了实现对于微博长文本谣言的检测,使用ERNIE模型对微博长文本数据使用词嵌入技术实现对文本的向量化,将这些词向量作为BiLSTM-DPCNN模型的输入,利用BiLSTM-DPCNN模型的特性,捕捉句子中的上下文信息和长期依赖关系,最终高效地提取特征并进行分类.数据集通过五折交叉验证方法划分,通过设置多组对比实验,实验结果显示:ERNIE-BiLSTM-DPCNN模型的准确率达到98.52%,高于在同一数据集下的其他同组实验,证明该模型对于微博长文本谣言检测具有较好的效果.

Moho反射波噪声干涉方法及其在青藏高原东北缘地壳结构成像的应用初探

基于南北地震带北段布设的密集流动地震台阵的连续波形记录,采用台阵噪声干涉技术获得了青藏高原东北缘Moho面PmP反射波,利用已有的区域三维地壳速度模型,采用沉积层校正方法建立研究区域的三维地壳初始模型,进而对噪声互相关函数进行基准面校正,然后采用共中心点叠加方法获得了研究区的地壳厚度分布.结果显示,青藏高原东北缘及周边不同块体地壳厚度差异显著,沿青藏高原北东向的扩展方向,地壳厚度逐渐变薄.祁连造山带中、西段地壳厚度约55~65 km,西秦岭造山带地壳厚度从西至东从53 km左右减小至42 km上下,阿拉善块体地壳厚度约50 km,鄂尔多斯块体西缘地壳厚度约40~45 km.六盘山断裂、海原断裂与天景山断裂附近呈现局部的地壳增厚,并且出现脉冲状Moho面反射震相,表明断裂带附近可能存在速度跃变较大的Moho面结构.祁连山北缘断裂出现叠置的反射震相,推测下地壳的逆冲形成了此类Moho结构.

旋转盘离心雾化制备ErNi球形磁性蓄冷材料

采用旋转盘真空离心雾化设备制备了ErNi合金磁性蓄冷材料球形粉末。对粉末的形貌特征、组织结构以及物相组成进行了全面研究,结果表明,粉末为光滑规整的球形,粒度分布较为均匀,约95%左右的颗粒尺寸分布在600μm以下;颗粒由单一ErNi相组成,为近似等轴状晶粒结构,内部无孔洞;其氧含量为0.093%,碳含量为0.0080%,均在允许的杂质元素含量的范围之内,完全可以满足实际应用的需求。

基于ERNIE-BiLSTM的社交网络文本情感分析

社交网络文本情感分析任务中,因短文本信息模糊等特点,传统的词向量模型无法更好地表示词的语义特征,当前短文本情感分类任务多以二分类研究为主,将结果分类为积极情感与消极情感,未能对分类结果更细入的划分。文中提出一种舆情情感分析的ERNIE-BiLSTM方法,实现了对用户评论情感的七种情绪分类,包括恐惧、厌恶、乐观、惊喜、感恩、悲伤和愤怒。ERNIE-BiLSTM方法利用ERNIE预训练模型获取文本的语义信息,结合BiLSTM提取文本的双向特征,最后使用softmax函数获得最终的情感分类结果。实验结果表明,ERNIE-BiLSTM方法具有87.7%的精确率、86.9%的召回率和86.8%的F1得分,比其他方法得到了有效提升。

Impacts and Countermeasures of Climate Change on Animal Husbandry in Qinghai Plateau and Its Surrounding Area

The paper aims to study the impacts and countermeasures of climate change on animal husbandry in Qinghai plateau and its surrounding area,and evaluate the effect of countermeasures.Results showed that：First,the annual mean temperature,annual mean maximum temperature and annual mean minimum temperature showed an increasing trend in Qinghai plateau during the time from 1961 to 2008; the annual precipitation had not obvious change,but the obvious feature of seasonal change; wind speed and sunshine hours showed a decreasing trend during the time from 1961 to 2008.Secondly,under the current climate condition,grassland productivity showed a decreasing trend,and the grassland theoretical capacity affected by grassland productivity decreased too; warm winter was in favor of livestock overwintering safely,which resulted in a high level of the survival rate of young stock; temperature increasing and precipitation decreasing were beneficial to the fatness of livestock,but the meat yield of livestock was affected by some negative factors such as grassland degradation; livestock diseases had a close relationship with weather condition,and adverse weather condition may arose many diseases of livestock.Thirdly,in view of the current problems of livestock production,we took a series of countermeasures including grass industry construction,livestock improvement,developing new modes of animal husbandry,control and prevention of stockbreeding diseases and insect pests and recovery of meteorological disasters,which played a important role in promoting grassland productivity,solving the problem of the imbalances between livestock and forage,protecting the ecological environment and improving livestock performance.