基于RoBERTa-Span-Attack的标签指针网络军事命名实体识别

军事领域文本中存在大量军事实体信息,准确识别这些信息是军事文本信息提取和构建军事知识图谱的基础性任务。首先,提出了一种基于RoBERTa预训练模型、跨度和对抗训练的标签指针网络的融合深度模型(RoBERTa-Span-Attack),用于中文军事命名实体识别;然后,采用了一种基于Span的标签指针网络,同时完成实体的起止位置和类别的识别任务;最后,在模型训练过程中加入对抗训练策略,通过添加一些扰动来生成对抗样本进行训练。在军事领域数据集上的实验结果表明:所提出的军事领域命名实体识别模型相较于BERT-CRF、BERT-Softmax和BERT-Span,在识别准确度上具有更优的效果。

基于对span的预判断和多轮分类的实体关系抽取

针对自然语言处理领域中的实体识别和关系抽取任务,提出一种对词元序列(Token Sequence,又称span)进行预测的模型Smrc。模型整体上利用BERT预训练模型作为编码器,另外包含实体预判断(Pej)、实体多轮分类(Emr)和关系多轮分类(Rmr)3个模块。Smrc模型通过Pej模块的初步判断及Emr模块的多轮实体分类来进行实体识别,再利用Rmr模块的多轮关系分类来判断实体对间的关系,进而完成关系抽取任务。在CoNLL04、SciERC和ADE 3个实验数据集上,Smrc模型的实体识别F1值分别达到89.67%,70.62%和89.56%,关系抽取F1值分别达到73.11%,51.03%和79.89%,相较之前在3个数据集上的最佳模型Spert,Smrc模型凭借实体预判断和实体及关系多轮分类,在2个子任务上其F1值分别提高了0.73%,0.29%,0.61%及1.64%,0.19%,1.05%,表明了该模型的有效性及其优势。

EMSS:一种基于Span匹配的中文实体抽取方法

基于Span(跨度)的实体抽取模型目前在英文数据集上取得了优异的效果,且已被证明跨度实体抽取比传统的序列标注实体抽取的效果更好.本文提出了一种基于跨度与拼接的中文命名实体抽取模型(EMSS),EMSS使用端到端的span抽取模型,文本经过BERT预训练模型进行字向量的编码,随后进入span抽取层枚举出所有可能的跨度,并加入跨度边界与跨度长度两种特征向量用于对跨度向量进行计算,最后进入跨度预测层进行实体标签的预测.同时提出了一种基于BIO格式的新标签标注方法,该标注方法不受模型与数据集领域限制,可以在不影响下游任务的情况下提高模型识别准确率.在Weibo、Resume、MSRA、OntoNotes4.0数据集上与当前主流的中文实体抽取模型进行对比实验.实验结果表明,提出的EMSS优于现有主流模型,均取得了7%左右的F1值提升.并将该方法应用到煤矿机电设备领域,解决煤矿机电设备领域的实体识别问题,在自制数据集上的实验证明本文的标注方法,不仅在中文实体上有效,而且对汉字、英文、数字结合的混合类型实体也有明显的效果.

On Spanning Wide Diameter of Graphs

A t-container Ct(u,v)is a set of t internally disjoint paths between two distinct vertices u and v in a graph G,i.e.,Ct(u,v)={P_(1),P_(2),···,Pt}.Moreover,if V(P_(1))∪V(P_(2))∪···∪V(Pt)=V(G)then Ct(u,v)is called a spanning t-container,denoted by C_(t)^(sc)(u,v).The length of C_(t)^(sc)(u,v)={P_(1),P_(2),···,Pt}is l(C_(t)^(sc)(u,v))=max{l(P_(i))|1≤i≤t}.A graph G is spanning t-connected if there exists a spanning t-container between any two distinct vertices u and v in G.Assume that u and v are two distinct vertices in a spanning t-connected graph G.Let D_(t)^(sc)(u,v)be the collection of all C_(t)^(sc)(u,v)’s.Define the spanning t-wide distance between u and v in G,d_(t)^(sc)(u,v)=min{l(C_(t)^(sc)(u,v))|C_(t)^(sc)(u,v)∈D_(t)^(sc)(u,v)},and the spanning t-wide diameter of G,D_(t)^(sc)(G)=max{d_(t)^(sc)(u,v)|u,v∈V(G)}.In particular,the spanning wide diameter of G is D_(κ)^(sc)(G),whereκis the connectivity of G.In the paper we provide the upper and lower bounds of the spanning wide diameter of a graph,and show that the bounds are best possible.We also determine the exact values of wide diameters of some well known graphs including Harary graphs and generalized Petersen graphs et al..

Effects of water and nitrogen rate on grain-filling characteristics under high-low seedbed cultivation in winter wheat

A high-efficiency mode of high-low seedbed cultivation(HLSC)has been listed as the main agricultural technology to increase land utilization ratio and grain yield in Shandong Province,China.However,limited information is available on the optimized water and nitrogen management for yield formation,especially the grain-filling process,under HLSC mode.A three-year field experiment with four nitrogen rates and three irrigation rates of HLSC was conducted to reveal the response of grain-filling parameters,grain weight percentage of spike weight(GPS),spike moisture content(SMC),and winter wheat yield to water and nitrogen rates.The four nitrogen rates were N1(360 kg ha^(-1) pure N),N2(300 kg ha^(-1) pure N),N3(240 kg ha^(-1) pure N),and N4(180 kg ha^(-1) pure N),respectively,and the three irrigation quotas were W1(120 mm),W2(90 mm),and W3(60 mm),respectively.Results showed that the determinate growth function generally performed well in simulating the temporal dynamics of grain weight(0.989<R^(2)<0.999,where R2 is the determination coefficient).The occurrence time of maximum filling rate(T_(max))and active grain-filling period(AGP)increased with the increase in the water or nitrogen rate,whereas the average grain-filling rate(G_(mean))had a decreasing trend.The final 1,000-grain weight(FTGW)increased and then decreased with the increase in the nitrogen rates and increased with the increase in the irrigation rates.The GPS and SMC had a highly significant quadratic polynomial relationship with grain weight and days after anthesis.Nitrogen,irrigation,and year significantly affected the T_(max),AGP,G_(mean),and FTGW.Particularly,the AGP and FTGW were insignificantly different between high seedbed(HLSC-H)and low seedbed(HLSC-L)across the water and nitrogen levels.Moreover,the moderate water and nitrogen supply was more beneficial for grain yield,as well as for spike number and grain number per hectare.The principal component analysis indicated that combining 240-300 kg N ha^(-1) and 90^(-1)20 mm irrigation quota could improve grain-filling efficiency and yield for the HLSC-cultivated winter wheat.

基于Span-ASTE的可解释性政府多对象态度识别模型研究

[目的/意义]在政府态度识别研究中,针对没有考虑到新闻文本中可能存在多个评价对象、政府态度识别结果可解释性不强等问题,基于Span-ASTE构建政府态度识别模型,提升态度识别效果。[方法/过程]首先,采用BERT提取词级别特征,基于跨度转化为跨度特征表示;然后,联合方面术语和观点术语提取任务提取评价对象和态度描述语,利用双通道跨度剪枝策略筛选得到对象和态度描述语候选池;最后,结合候选池中的对象和态度描述语计算得到态度极性结果。[结果/结论]以美国国务院新闻文本为实验数据进行实验验证。实验结果表明:Span-ASTE在进行政府态度识别时具有一定优越性。相比于效果较好的对比模型,其精确率、召回率和F1值分别提升了约15.68%、19.37%和17.48%;在进行多对象态度识别时同样具有良好的性能表现;态度描述语可为态度极性的判断提供解释依据。[局限]政府态度识别效果还有待进一步提升,且数据规模较为有限,未来可尝试对数据集进行扩充。

Analyzing the Form-Finding of a Large-Span Transversely Stiffened Suspended Cable System: A Method Considering Construction Processes

The precise control of the shape of transversely stiffened suspended cable systems is crucial. However, existing form-finding methods primarily rely on iterative calculations that treat loads as fixed known conditions. These methods are inefficient and fail to accurately control shape results. In this study, we propose a form-finding method that analyzes the load response of models under different sag and stress levels, taking into account the construction process. To analyze the system, a structural finite element model was established in ANSYS, and geometric nonlinear analysis was conducted using the Newton-Raphson method. The form-finding analysis results demonstrate that the proposed method achieves precise control of shape, with a maximum shape error ranging from 0.33% to 0.98%. Furthermore, the relationships between loads and tension forces are influenced by the deformed shape of the structures, exhibiting significant geometric nonlinear characteristics. Meanwhile, the load response analysis reveals that the stress level of the self-equilibrium state in the transversely stiffened suspended cable system is primarily governed by strength criteria, while shape is predominantly controlled by stiffness criteria. Importantly, by simulating the initial tensioning process as an initial condition, this method solves for a counterweight that satisfies the requirements and achieves a self-equilibrium state with the desired shape. The shape of the self-equilibrium state is precisely controlled by simulating the construction process. Overall, this work presents a new method for analyzing the form-finding process of large-span transversely stiffened suspended cable system, considering the construction process which was often overlooked in previous studies.

探讨SPAN指数、wSPAN指数和SC评分对急性缺血性卒中血管内治疗患者的临床预后

目的探讨术前年龄和NIHSS两者相结合的指标SPAN指数、wSPAN指数、SC评分对急性缺血性卒中(AIS)患者血管内治疗术后临床预后之间的关系。方法连续收集2015年8月—2021年10月就诊于中国3个综合性卒中中心接受血管内治疗的AIS患者889例,以门诊或电话方式随访90天。根据术后90天改良Rankin量表(90 d mRS)评分评估患者预后,分为预后良好组(mRS≤2分)和预后不良组(mRS>2分)。同时记录患者术后症状性颅内出血(sICH)、颅内出血转化的发生率及术后90天死亡率。结果在纳入的889例患者中,预后不良组血管内治疗前的SPAN指数、wSPAN指数、SC评分均显著高于预后良好组(P<0.001)。在调整潜在的混杂因素后,SPAN、wSPAN、SC均与EVT患者90天预后独立相关(SPAN:OR=0.961,95%CI=0.948-0.975、wSPAN:OR=0.968,95%CI=0.959-0.976、SC:OR=0.767,95%CI=0.713-0.826)。SPAN、wSPAN、SC与sICH之间没有统计学差异(P=0.220,P=0.088,P=0.072),而与颅内出血转化风险增加(SPAN:OR=1.03,95%CI=1.02-1.05,P<0.001、wSPAN:OR=1.03,95%CI=1.02-1.03,P<0.001、SC:OR=1.25,95%CI=1.14-1.36,P<0.001)和死亡率(P<0.001,P<0.001,P<0.001)相关。结论术前SPAN指数、wSPAN指数、SC评分与AIS患者血管内治疗90天临床预后、颅内出血转化发生率和高死亡率相关,但与sICH无关。

基于改进Prefix Span的结构超手册修理方法推荐

针对当前民机结构超手册修理能力提升的迫切需求,提出了一种基于改进Prefix Span的结构超手册修理方法推荐模型。以某型国产民机结构超手册修理案例数据为样本,构建案例特征指标层次并对数据分级和结构化预处理,基于改进Prefix Span算法对案例损伤信息和修理方法频繁序列进行充分挖掘,生成超手册修理关联规则集,通过目标案例和频繁序列间相似度匹配,实现了超手册修理方法的智能推荐。测试结果表明,模型超手册修理方法推荐总准确率可达85.3%,显著高于改进前算法,具备良好的准确性和可用性,可为结构超手册修理决策、方案制定等提供一定的参考依据。

Brain Functional Network Based on Small-Worldness and Minimum Spanning Tree for Depression Analysis

Since the outbreak and spread of corona virus disease 2019(COVID-19),the prevalence of mental disorders,such as depression,has continued to increase.To explore the abnormal changes of brain functional connections in patients with depression,this paper proposes a depression analysis method based on brain function network(BFN).To avoid the volume conductor effect,BFN was constructed based on phase lag index(PLI).Then the indicators closely related to depression were selected from weighted BFN based on small-worldness(SW)characteristics and binarization BFN based on the minimum spanning tree(MST).Differences analysis between groups and correlation analysis between these indicators and diagnostic indicators were performed in turn.The resting state electroencephalogram(EEG)data of 24 patients with depression and 29 healthy controls(HC)was used to verify our proposed method.The results showed that compared with HC,the information processing of BFN in patients with depression decreased,and BFN showed a trend of randomization.

A method for calculating strand tension in the anchor span of a suspension bridge considering the rotation of a splay saddle

This paper reports a method for strand tension in anchor spans considering rotation.A kind of co-moved coordinate system,a saddle local coordinate system,was set up.This system implemented the rotation of the splay saddle through the rotation of the coordinate system,and all calculations proceeded in this coordinate system.Considering the rotation of the anchoring surface by the rotation of the local coordinate system of the anchoring surface,the anchorage point coordinates of strands were transformed to the local sadle coordinate system.There was a two-layer iteration adopted in the calculation.In the inner iteration,the cable force at the end of the vertical bend was taken as the variable,and the ordinate of the anchorage point was taken as the target value.In the outer iteration,the vertical tangential angle at the end of the vertical bend was taken as the variable,and the ordinate of the anchorage point was taken as the target value.The method carried out the rotation of the splay saddle and anchor surface and was simple,convenient and without approximation.The effect of rotation was considered precisely;it showed stability during the process of two-layer iteration,powerful adaptation and higher efficiency and had been successfully applied in the construction control of the Wufengshan Yangtze River Bridge,the world's first kilometer-level combined highway and railway suspension bridge.

Stability evaluation method of large cross-section tunnel considering modification of thickness-span ratio in mechanized operation

Purpose-This study aims to research the large cross-section tunnel stability evaluation method corrected after considering the thickness-span ratio.Design/methodology/approach-First,taking the Liuyuan Tunnel of Huanggang-Huangmei High-Speed Railway as an example and taking deflection of the third principal stress of the surrounding rock at a vault after tunnel excavation as the criterion,the critical buried depth of the large section tunnel was determined.Then,the strength reduction method was employed to calculate the tunnel safety factor under different rock classes and thickness-span ratios,and mathematical statistics was conducted to identify the relationships of the tunnel safety factor with the thickness-span ratio and the basic quality(BQ)index of the rock for different rock classes.Finally,the influences of thickness-span ratio,groundwater,initial stress of rock and structural attitude factors were considered to obtain the corrected BQ,based on which the stability of a large cross-section tunnel with a depth of more than 100 m during mechanized operation was analyzed.This evaluation method was then applied to Liuyuan Tunnel and Cimushan No.2 Tunnel of Chongqing Urban Expressway for verification.Findings-This study shows that under different rock classes,the tunnel safety factor is a strict power function of the thickness-span ratio,while a linear function of the BQ to some extent.It is more suitable to use the corrected BQ as a quantitative index to evaluate tunnel stability according to the actual conditions of the site.Originality/value-The existing industry standards do not consider the influence of buried depth and span in the evaluation of tunnel stability.The stability evaluation method of large section tunnel considering the correction of overburden span ratio proposed in this paper achieves higher accuracy for the stability evaluation of surrounding rock in a full or large-section mechanized excavation of double line high-speed railway tunnels.

深井松软围岩煤巷采动增跨效应及防控技术

针对深井松软煤巷围岩变形严重、巷道支护困难等问题,以城郊煤矿LW21106工作面沿空巷道为工程背景,建立了采动巷道增跨模型,揭示了采动增跨效应演化机理。通过构建巷道顶板横纵弯曲梁模型,指出顶板横向受力、巷道等效跨度、煤岩强度是巷道围岩损伤破坏的主控因素,提出了采动增跨效应防控对策并进行工业性试验。研究结果表明:受采动影响,巷道经历“初始围岩稳定—围岩裂隙发育扩展—围岩剪切破坏加剧—等效跨度增加”过程;巷道顶板最大正应力与应力集中系数、顶板等效跨度、巷道断面尺寸及埋深成正相关关系;巷道顶板在高应力环境下易发生拉剪破坏,增加顶板锚索数量以及锚索预紧力有利于增强顶板初期完整性。基于巷道变形破坏主控因素,提出“围岩加固–卸压–强化支护”协同防控策略;针对现场条件,采用煤柱侧向切顶+注浆加固并对破碎区域补充锚索强化支护的防控技术。现场监测结果表明,煤柱帮最大移近量为18.89cm,顶板下沉量为25.86cm,两帮移近量为29.65 cm,有效控制了煤巷围岩变形,为深井松软围岩巷道变形控制提供了参考。

复杂环境下秀江双桥控制爆破拆除

为在复杂环境下对两座200 m钢筋砼桥梁(连续梁老桥与拱形新桥)实施控制爆破,选择双桥桥墩、连续梁腹板及双曲拱桥拱脚为爆破部位,破坏桥梁的主要承重部位。采用深孔桥墩爆破和浅孔主梁爆破相结合的爆破方法,降低施工安全风险并提高施工效率。其中,深孔桥墩爆破采用由桥面向桥墩的大直径垂直炮孔,克服了桥墩圆弧面对最小抵抗线计算的影响,减少了钻孔工作量。设计起爆方式为梁桥同方向逐跨分段延期起爆,即:两桥由南向北同向逐跨起爆,每跨之间等间隔延时,同时两桥之间设计50 ms的延时间隔(新桥的起爆时间比老桥延后50 ms)。为减少爆破振动及塌落振动,采用毫秒延时控制爆破,最小抵抗线方向背离较近岸边保护对象。此外,为控制爆破飞石,采用了桥台堆沙袋,爆破体进行钢丝网捆扎并用多层竹笆覆盖的防护措施。起爆后,大桥逐跨坍塌,被爆桥体塌落解体较充分,塌落振动及飞石控制效果良好。

大跨度钢桁架结构卸载模拟及监测技术研究

以江苏省盐城黄淮湿地博物馆钢桁架结构卸载施工为背景,旨在通过有限元技术及基于多传感器融合的现场监测技术研究大跨度桁架结构卸载施工关键技术。研究结果表明:有限元计算结果能够获取卸载施工过程中的关键要素,为后续监测方案的设计提供参考。后续的监测结果表明,有限元方法对位移的计算结果可靠性要高于应力,与实际情况更加接近;基于多传感器融合的监测方法能够提高监测结果的准确性和远程监控效果,为桁架结构的安全性和可靠性提供了有力支撑。研究结果可为大跨度钢结构建筑的施工提供参考。

大跨度斜拉桥模态参数长期追踪及其变异性分析

为了丰富大跨度斜拉桥模态参数的实测数据库,依托苏通大桥结构健康监测系统采集的数据,采用所建立的模态自动识别和追踪方法,获取了2010年期间该桥的模态参数值,并据此分析了在温度和风速作用下桥梁模态参数的变异性。研究结果表明,大桥主梁模态频率的变化受温度和风速共同影响,随着温度的升高而降低,随着风速的升高而增加。大桥主梁模态阻尼比的变异性显著大于模态频率,低风速下主梁一阶侧弯阻尼比在0.5%~15%之间波动,风速大于9 m/s时逐渐降低并稳定在2%左右。大桥主梁前四阶竖弯阻尼比受气动阻尼的影响较大,在风速较小时随着风速的增加略有增加。研究结果可为大桥的服役性能评估与运营管理提供参考。

公铁两用大跨度斜拉桥-无砟轨道体系变形适应性研究

高速铁路大跨度斜拉桥结构变形受其复杂服役环境影响显著,而无砟轨道对工后变形的调整能力有限,因此大跨度斜拉桥的复杂变形条件将直接影响其上无砟轨道结构的适应性。以沪渝蓉高铁长江北支公铁两用大跨度斜拉桥应用无砟轨道为工程背景,建立无砟轨道-大跨度斜拉桥一体化精细空间有限元模型,分析铁路、公路以及温度等多种荷载联合作用下大跨度斜拉桥-无砟轨道体系平顺性、无缝线路稳定性和无砟轨道层间变形协调性,从而对大跨度公铁两用钢桁梁斜拉桥-无砟轨道体系变形适应性开展深入研究。结果表明:在运营荷载及温度荷载最不利组合作用下,大跨度斜拉桥主梁线形平顺,变形平缓,桥梁整体刚度较好,桥上无砟轨道几何形位可以满足静态验收标准;大跨度斜拉桥上无缝线路的强度和稳定性满足要求,扣件工作性能良好,钢轨伸缩调节器的设置可有效释放梁端钢轨的纵向应力;无砟轨道层间设置橡胶隔离层且不跨主梁桁架节间的布置方式可提高大跨度斜拉桥上无砟轨道的变形协调性能,使得桥上无砟轨道层间始终处于受压状态,达到“隔而不离”的桥上轨道层间理想服役状态;大跨度公铁两用钢桁梁斜拉桥-无砟轨道体系变形适应性良好。本文研究成果可为公铁两用大跨度钢桁梁斜拉桥上无砟轨道的应用提供技术参考。

大跨越架空输电导线铝部应力分布试验

为了掌握大跨越架空输电导线铝部应力空间分布特性,以JLHA1/G6A-500/280型特强钢芯铝合金导线为研究对象,采用铝股丝表面激光刻槽,内嵌高灵敏、超大复用容量超弱光纤光栅,搭建了大跨越输电导线铝部应力试验平台,研究了不同张力条件下大跨越架空输电导线铝部应力分布规律,建立了大跨越导线实体有限元模型,分析了铝股应力分布特征,并验证了试验结果的正确性。结果表明:导线承受张力作用时,外层铝股和次外层铝股应力不同,外层铝股应力小于次外层铝股应力,导线同层铝股应力基本相同;铝股应力随导线张力的增大呈线性增大趋势,张力每增大1%,铝股应力增大约10%。不同导线张力作用时钢股和铝股应力均呈现环状分层特性,铝股平均应力小于钢股平均应力,铝部和钢股应力比约为3∶7;外层和次外层铝合金股丝应力沿圆周向分布不均匀,建议大跨越导线线型设计时考虑铝股丝分层特性。

考虑结构空间约束的现浇楼盖梁跨中受弯性能试验

为研究结构空间约束导致现浇钢筋混凝土楼盖梁跨中受弯承载力超强的机制,以楼盖梁在结构中的空间位置和配筋率为变量,设计12根上部架立筋不伸入支座的楼盖梁试件并开展跨中受弯静力试验,并基于建立经典混凝土理论的试验方法进行了4根简支梁对比试验。对梁试件破坏形态、承载力、变形能力、轴向伸长进行研究,结果表明:楼盖梁试件承载力与按受弯构件计算承载力之比为1.57~2.77,与相应矩形简支梁试件和T形简支梁试件承载力相比均有较大提高,超强程度与配筋率成反比;与简支梁试件相比,楼盖梁试件跨中截面破坏形态未因轴压力而发生显著改变,但楼盖梁试件延性降低;楼盖梁试件轴向伸长随挠度的增大而增大。将楼盖梁视为压弯构件,提出根据楼盖梁试验峰值荷载确定相应轴力和弯矩的方法。计算结果表明,楼盖梁试件跨中截面压弯状态弯矩与受弯状态弯矩的比值为1.28~2.19。楼盖梁试件有限元数值模拟结果表明,随着楼盖梁截面包含的楼板范围增大,轴力沿梁跨分布趋于均匀。

中文阅读中快速读者与慢速读者的知觉广度

在熟练的中文阅读中,不同阅读速度读者提取文本信息的范围(即知觉广度)是否存在差异?为解决该问题,本研究从熟练读者群体中选取快速读者与慢速读者,结合眼动技术和移动窗口范式比较两组读者在不同窗口(R0、L1R1、L1R2、L1R3、L1R4、FL)下阅读句子时的眼动模式。结果显示,在保证阅读理解准确的前提下,快速读者比慢速读者的知觉广度更大,符合E-Z读者模型的语言效率假说,支持个体差异性观点。