融合多尺度CNN和CRF的通用细粒度事件检测

事件检测是自然语言处理领域中事件抽取的主要任务之一,它旨在从众多非结构化信息中自动提取出结构化的关键信息.现有的方法存在特征提取不全面、特征分布不均等情况.为了提高事件检测的准确率,提出了一种融合BERT预训练模型与多尺度CNN的神经网络模型(BMCC,BERT+Multi-scale CNN+CRF).首先通过BERT(Bidirectional Encoder Representations from Transformers)预训练模型来进行词向量的嵌入,并利用其双向训练的Transformer机制来提取序列的状态特征;其次使用不同尺度的卷积核在多个卷积通道中进行卷积训练,以此来提取不同视野的语义信息,丰富其语义表征.最后将BIO机制融入到条件随机场(CRF)来对序列进行标注,实现事件的检测.实验结果表明,所提出的模型在MAVEN数据集上的F1值为65.17%,表现了该模型的良好性能.

基于RoBERTa_BiLSTM_CRF的文本情报命名实体识别

随着网络信息的爆炸式增长,威胁情报分析作为军事情报分析与战略决策的重要组成部分,其面临着来源多样化和信息结构复杂化的挑战。传统的人工信息提取方法在处理这些大量结构化及非结构化信息时效率低下,准确性有限。文中针对这一挑战,提出了一种结合RoBERTa、BiLSTM和条件随机场(Conditional Random Fields,CRF)的命名实体识别新算法。此算法通过Ro-BERTa模型深入挖掘文本的语义特征,BiLSTM模型捕捉序列上下文信息,CRF层用于精确的实体标记,从而有效提升信息提取的准确率和效率。本文基于开源情报语料库构建了一个涉及导弹发射事件的命名实体识别数据集,并在此基础上进行了实验,结果表明,该方法在精确率、召回率及F1值等关键指标上相较于主流深度学习方法表现出显著的性能提升,其中F1值高达94.21%。

基于BERT-BiLSTM-CRF模型的畜禽疫病文本分词研究

针对畜禽疫病文本语料匮乏、文本内包含大量疫病名称及短语等未登录词问题,提出了一种结合词典匹配的BERT-BiLSTM-CRF畜禽疫病文本分词模型。以羊疫病为研究对象,构建了常见疫病文本数据集,将其与通用语料PKU结合,利用BERT(Bidirectional encoder representation from transformers)预训练语言模型进行文本向量化表示;通过双向长短时记忆网络(Bidirectional long short-term memory network,BiLSTM)获取上下文语义特征;由条件随机场(Conditional random field,CRF)输出全局最优标签序列。基于此,在CRF层后加入畜禽疫病领域词典进行分词匹配修正,减少在分词过程中出现的疫病名称及短语等造成的歧义切分,进一步提高了分词准确率。实验结果表明,结合词典匹配的BERT-BiLSTM-CRF模型在羊常见疫病文本数据集上的F1值为96.38%,与jieba分词器、BiLSTM-Softmax模型、BiLSTM-CRF模型、未结合词典匹配的本文模型相比,分别提升11.01、10.62、8.3、0.72个百分点,验证了方法的有效性。与单一语料相比,通用语料PKU和羊常见疫病文本数据集结合的混合语料,能够同时对畜禽疫病专业术语及疫病文本中常用词进行准确切分,在通用语料及疫病文本数据集上F1值都达到95%以上,具有较好的模型泛化能力。该方法可用于畜禽疫病文本分词。

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BERT-BiLSTM-CRF模型的油气领域命名实体识别

针对油气领域知识图谱构建过程中命名实体识别使用传统方法存在实体特征信息提取不准确、识别效率低的问题,提出了一种基于BERT-BiLSTM-CRF模型的命名实体识别研究方法。该方法首先利用BERT(bidirectional encoder representations from transformers)预训练模型得到输入序列语义的词向量;然后将训练后的词向量输入双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)模型进一步获取上下文特征;最后根据条件随机场(conditional random fields,CRF)的标注规则和序列解码能力输出最大概率序列标注结果,构建油气领域命名实体识别模型框架。将BERT-BiLSTM-CRF模型与其他2种命名实体识别模型(BiLSTM-CRF、BiLSTM-Attention-CRF)在包括3万多条文本语料数据、4类实体的自建数据集上进行了对比实验。实验结果表明,BERT-BiLSTM-CRF模型的准确率(P)、召回率(R)和F_(1)值分别达到91.3%、94.5%和92.9%,实体识别效果优于其他2种模型。

基于改进BiLSTM-CRF模型的网络安全知识图谱构建

针对网络安全领域的图谱构建任务,基于BiLSTM-CRF模型引入了外部网络安全词典来加强网络安全文本的特征,并结合多头注意力机制提取多层特征,最终在网络安全数据集取得了更优异的结果。利用企业内部的日常网络运维数据,设计并构建了一个面向企业网络安全运维管理的知识图谱,为后续进一步研究基于图谱的企业网络安全智能决策等应用奠定基础。

Bi⁃LSTM和CRF结合的藏文分词方法研究

本研究旨在探索基于双向长短时记忆网络(Bi⁃LSTM)和条件随机场(CRF)的藏语分词方法。研究评估了Bi⁃LSTM和Bi⁃LSTM结合CRF模型在藏文分词任务中的性能,针对未登录音节词进行了模型优化。研究还将分词系统应用到CCMT2023藏汉机器翻译任务中,并进行了评估。实验结果表明,本文提出的分词系统在藏语分词任务中具有良好的性能,并且在藏汉机器翻译任务上有效提升了性能。

基于BERT-BiLSTM-CRF模型的中医治疗功能性胃肠病实体识别及应用

目的:探索分析BERT-BiLSTM-CRF模型抽取中医文献摘要中的实体的可行性及识别效果。方法:在知网数据中导出500条中医疗法治疗功能性胃肠病的论文摘要,对文本中的西医病名、临床表现、方剂、中药等11类实体进行BIO标注,基于BERT-BiLSTM-CRF模型进行训练及参数调整,而后对模型进行测试,并应用于实体识别。结果:模型测试的精确率为85.07%,召回率为88.48%,F1值为0.8674,中药、方剂、西医诊断等实体类别的识别效果较好;模型应用中,自动化实体抽取结果整体较好,能够反映该领域文献的主要研究方向。结论:BERT-BiLSTM-CRF模型能够识别出论文摘要中大部分的实体,可以为知识图谱的自动化构建提供基础,同时也对中医药领域的自然语言处理应用提供了参考和借鉴。

Drilling-based measuring method for the c-φ parameter of rock and its field application

The technology of drilling tests makes it possible to obtain the strength parameter of rock accurately in situ. In this paper, a new rock cutting analysis model that considers the influence of the rock crushing zone(RCZ) is built. The formula for an ultimate cutting force is established based on the limit equilibrium principle. The relationship between digital drilling parameters(DDP) and the c-φ parameter(DDP-cφ formula, where c refers to the cohesion and φ refers to the internal friction angle) is derived, and the response of drilling parameters and cutting ratio to the strength parameters is analyzed. The drillingbased measuring method for the c-φ parameter of rock is constructed. The laboratory verification test is then completed, and the difference in results between the drilling test and the compression test is less than 6%. On this basis, in-situ rock drilling tests in a traffic tunnel and a coal mine roadway are carried out, and the strength parameters of the surrounding rock are effectively tested. The average difference ratio of the results is less than 11%, which verifies the effectiveness of the proposed method for obtaining the strength parameters based on digital drilling. This study provides methodological support for field testing of rock strength parameters.

Strong coupling and catenary field enhancement in the hybrid plasmonic metamaterial cavity and TMDC monolayers

Strong coupling between resonantly matched surface plasmons of metals and excitons of quantum emitters results in the formation of new plasmon-exciton hybridized energy states.In plasmon-exciton strong coupling,plasmonic nanocavities play a significant role due to their ability to confine light in an ultrasmall volume.Additionally,two-dimensional transition metal dichalcogenides(TMDCs) have a significant exciton binding energy and remain stable at ambient conditions,making them an excellent alternative for investigating light-matter interactions.As a result,strong plasmon-exciton coupling has been reported by introducing a single metallic cavity.However,single nanoparticles have lower spatial confinement of electromagnetic fields and limited tunability to match the excitonic resonance.Here,we introduce the concept of catenary-shaped optical fields induced by plasmonic metamaterial cavities to scale the strength of plasmon-exciton coupling.The demonstrated plasmon modes of metallic metamaterial cavities offer high confinement and tunability and can match with the excitons of TMDCs to exhibit a strong coupling regime by tuning either the size of the cavity gap or thickness.The calculated Rabi splitting of Au-MoSe_2 and Au-WSe_2 heterostructures strongly depends on the catenary-like field enhancement induced by the Au cavity,resulting in room-temperature Rabi splitting ranging between 77.86 and 320 me V.These plasmonic metamaterial cavities can pave the way for manipulating excitons in TMDCs and operating active nanophotonic devices at ambient temperature.

Displacement field reconstruction in landslide physical modeling by using a terrain laser scanner e Part 2:Application and large strain/displacement and water effect analysis

Deformation analysis is fundamental in geotechnical modeling.Nevertheless,there is still a lack of an effective method to obtain the deformation field under various experimental conditions.In this study,we introduce a processebased physical modeling of a pileereinforced reservoir landslide and present an improved deformation analysis involving large strains and water effects.We collect multieperiod point clouds using a terrain laser scanner and reconstruct its deformation field through a point cloud processing workflow.The results show that this method can accurately describe the landslide surface deformation at any time and area by both scalar and vector fields.The deformation fields in different profiles of the physical model and different stages of the evolutionary process provide adequate and detailed landslide information.We analyze the large strain upstream of the pile caused by the pile installation and the consequent violent deformation during the evolutionary process.Furthermore,our method effectively overcomes the challenges of identifying targets commonly encountered in geotechnical modeling where water effects are considered and targets are polluted,which facilitates the deformation analysis at the wading area in a reservoir landslide.Eventually,combining subsurface deformation as well as numerical modeling,we comprehensively analyze the kinematics and failure mechanisms of this complicated object involving landslides and pile foundations as well as water effects.This method is of great significance for any geotechnical modeling concerning large-strain analysis and water effects.

ST-LSTM-SA:A New Ocean Sound Velocity Field Prediction Model Based on Deep Learning

The scarcity of in-situ ocean observations poses a challenge for real-time information acquisition in the ocean.Among the crucial hydroacoustic environmental parameters,ocean sound velocity exhibits significant spatial and temporal variability and it is highly relevant to oceanic research.In this study,we propose a new data-driven approach,leveraging deep learning techniques,for the prediction of sound velocity fields(SVFs).Our novel spatiotemporal prediction model,STLSTM-SA,combines Spatiotemporal Long Short-Term Memory(ST-LSTM) with a self-attention mechanism to enable accurate and real-time prediction of SVFs.To circumvent the limited amount of observational data,we employ transfer learning by first training the model using reanalysis datasets,followed by fine-tuning it using in-situ analysis data to obtain the final prediction model.By utilizing the historical 12-month SVFs as input,our model predicts the SVFs for the subsequent three months.We compare the performance of five models:Artificial Neural Networks(ANN),Long ShortTerm Memory(LSTM),Convolutional LSTM(ConvLSTM),ST-LSTM,and our proposed ST-LSTM-SA model in a test experiment spanning 2019 to 2022.Our results demonstrate that the ST-LSTM-SA model significantly improves the prediction accuracy and stability of sound velocity in both temporal and spatial dimensions.The ST-LSTM-SA model not only accurately predicts the ocean sound velocity field(SVF),but also provides valuable insights for spatiotemporal prediction of other oceanic environmental variables.

Field test of high-power microwave-assisted mechanical excavation for deep hard iron ore

Microwave-assisted mechanical excavation has great application prospects in mines and tunnels,but there are few field experiments on microwave-assisted rock breaking.This paper takes the Sishanling iron mine as the research object and adopts the self-developed high-power microwave-induced fracturing test system for hard rock to conduct field experiments of microwave-induced fracturing of iron ore.The heating and reflection evolution characteristics of ore under different microwave parameters(antenna type,power,and working distance)were studied,and the optimal microwave parameters were obtained.Subsequently,the ore was irradiated with the optimal microwave parameters,and the cracking effect of the ore under the action of the high-power open microwave was analyzed.The results show that the reflection coefficient(standing wave ratio)can be rapidly(<5 s)and automatically adjusted below the preset threshold value(1.6)as microwave irradiation is performed.When using a right-angle horn antenna with a working distance of 5 cm,the effect of automatic reflection adjustment reaches the best among other antenna types and working distances.When the working distance is the same,the average temperature of the irradiation surface and the area of the high-temperature area under the action of the two antennas(right-angled and equal-angled horn antenna)are basically the same and decrease with the increase of working distance.The optimal microwave parameters are:a right-angle horn antenna with a working distance of 5 cm.Subsequently,in further experiments,the optimal parameters were used to irradiate for 20 s and 40 s at a microwave power of 60 kW,respectively.The surface damage extended 38 cm×30 cm and 53 cm×30 cm,respectively,and the damage extended to a depth of about 50 cm.The drilling speed was increased by 56.2%and 66.5%,respectively,compared to the case when microwaves were not used.

Reactor field reconstruction from sparse and movable sensors using Voronoi tessellation-assisted convolutional neural networks

The aging of operational reactors leads to increased mechanical vibrations in the reactor interior.The vibration of the incore sensors near their nominal locations is a new problem for neutronic field reconstruction.Current field-reconstruction methods fail to handle spatially moving sensors.In this study,we propose a Voronoi tessellation technique in combination with convolutional neural networks to handle this challenge.Observations from movable in-core sensors were projected onto the same global field structure using Voronoi tessellation,holding the magnitude and location information of the sensors.General convolutional neural networks were used to learn maps from observations to the global field.The proposed method reconstructed multi-physics fields(including fast flux,thermal flux,and power rate)using observations from a single field(such as thermal flux).Numerical tests based on the IAEA benchmark demonstrated the potential of the proposed method in practical engineering applications,particularly within an amplitude of 5 cm around the nominal locations,which led to average relative errors below 5% and 10% in the L_(2) and L_(∞)norms,respectively.

Electromagnetic fields in ultra-peripheral relativistic heavy-ion collisions

Ultra-peripheral heavy-ion collisions(UPCs)offer unique opportunities to study processes under strong electromagnetic fields.In these collisions,highly charged fast-moving ions carry strong electromagnetic fields that can be effectively treated as photon fluxes.The exchange of photons can induce photonuclear and two-photon interactions and excite ions.This excitation of the ions results in Coulomb dissociation with the emission of photons,neutrons,and other particles.Additionally,the electromagnetic fields generated by the ions can be sufficiently strong to enforce mutual interactions between the two colliding ions.Consequently,the two colliding ions experience an electromagnetic force that pushes them in opposite directions,causing a back-to-back correlation in the emitted neutrons.Using a Monte Carlo simulation,we qualitatively demonstrate that the above electromagnetic effect is large enough to be observed in UPCs,which would provide a clear means to study strong electromagnetic fields and their effects.

Constructing Built-In Electric Fields with Semiconductor Junctions and Schottky Junctions Based on Mo-MXene/Mo-Metal Sulfides for Electromagnetic Response

The exploration of novel multivariate heterostructures has emerged as a pivotal strategy for developing high-performance electromagnetic wave(EMW)absorption materials.However,the loss mechanism in traditional heterostructures is relatively simple,guided by empirical observations,and is not monotonous.In this work,we presented a novel semiconductor-semiconductor-metal heterostructure sys-tem,Mo-MXene/Mo-metal sulfides(metal=Sn,Fe,Mn,Co,Ni,Zn,and Cu),including semiconductor junctions and Mott-Schottky junctions.By skillfully combining these distinct functional components(Mo-MXene,MoS_(2),metal sulfides),we can engineer a multiple heterogeneous interface with superior absorption capabilities,broad effective absorption bandwidths,and ultrathin matching thickness.The successful establishment of semiconductor-semiconductor-metal heterostructures gives rise to a built-in electric field that intensifies electron transfer,as confirmed by density functional theory,which collaborates with multiple dielectric polarization mechanisms to substantially amplify EMW absorption.We detailed a successful synthesis of a series of Mo-MXene/Mo-metal sulfides featuring both semiconductor-semiconductor and semiconductor-metal interfaces.The achievements were most pronounced in Mo-MXene/Mo-Sn sulfide,which achieved remarkable reflection loss values of-70.6 dB at a matching thickness of only 1.885 mm.Radar cross-section calculations indicate that these MXene/Mo-metal sulfides have tremendous potential in practical military stealth technology.This work marks a departure from conventional component design limitations and presents a novel pathway for the creation of advanced MXene-based composites with potent EMW absorption capabilities.

Enhancing the resolution of sparse rock property measurements using machine learning and random field theory

The travel time of rock compressional waves is an essential parameter used for estimating important rock properties,such as porosity,permeability,and lithology.Current methods,like wireline logging tests,provide broad measurements but lack finer resolution.Laboratory-based rock core measurements offer higher resolution but are resource-intensive.Conventionally,wireline logging and rock core measurements have been used independently.This study introduces a novel approach that integrates both data sources.The method leverages the detailed features from limited core data to enhance the resolution of wireline logging data.By combining machine learning with random field theory,the method allows for probabilistic predictions in regions with sparse data sampling.In this framework,12 parameters from wireline tests are used to predict trends in rock core data.The residuals are modeled using random field theory.The outcomes are high-resolution predictions that combine both the predicted trend and the probabilistic realizations of the residual.By utilizing unconditional and conditional random field theories,this method enables unconditional and conditional simulations of the underlying high-resolution rock compressional wave travel time profile and provides uncertainty estimates.This integrated approach optimizes the use of existing core and logging data.Its applicability is confirmed in an oil project in West China.

Flow field, sedimentation, and erosion characteristics around folded linear HDPE sheet sand fence: Numerical simulation study

Wind and sand hazards are serious in the Milan Gobi area of the Xinjiang section of the Korla Railway. In order to ensure the safe operation of railroads, there is a need for wind and sand protection in heavily sandy areas. The wind and sand flow in the region is notably bi-directional. To shield railroads from sand, a unique sand fence made of folded linear high-density polyethylene(HDPE) is used, aligning with the principle that the dominant wind direction is perpendicular to the fence. This study employed field observations and numerical simulations to investigate the effectiveness of these HDPE sand fences in altering flow field distribution and offering protection. It also explored how these fences affect the deposition and erosion of sand particles. Findings revealed a significant reduction in wind speed near the fence corner;the minimum horizontal wind speed on the leeward side of the first sand fence(LSF) decreased dramatically from 3 m/s to 0.64 m/s. The vortex area on the LSF markedly impacted horizontal wind speeds. Within the LSF, sand deposition was a primary occurrence. As wind speeds increased, the deposition zone shrank, whereas the positive erosion zone expanded. Close to the folded corners of the HDPE sand fence, there was a notable shift from the positive erosion zone to a deposition zone. Field tests and numerical simulations confirmed the high windproof efficiency(WE) and sand resistance efficiency(SE) in the HDPE sand fence. Folded linear HDPE sheet sand fence can effectively slow down the incoming flow and reduce the sand content, thus achieving good wind and sand protection. This study provides essential theoretical guidance for the design and improvement of wind and sand protection systems in railroad engineering.

Subsoil tillage enhances wheat productivity,soil organic carbon and available nutrient status in dryland fields

Tillage practices during the fallow period benefit water storage and yield in dryland wheat crops.However,there is currently no clarity on the responses of soil organic carbon(SOC),total nitrogen(TN),and available nutrients to tillage practices within the growing season.This study evaluated the effects of three tillage practices(NT,no tillage;SS,subsoil tillage;DT,deep tillage)over five years on soil physicochemical properties.Soil samples at harvest stage from the fifth year were analyzed to determine the soil aggregate and aggregate-associated C and N fractions.The results indicated that SS and DT improved grain yield,straw biomass and straw carbon return of wheat compared with NT.In contrast to DT and NT,SS favored SOC and TN concentrations and stocks by increasing the soil organic carbon sequestration rate(SOCSR)and soil nitrogen sequestration rate(TNSR)in the 0-40 cm layer.Higher SOC levels under SS and NT were associated with greater aggregate-associated C fractions,while TN was positively associated with soluble organic nitrogen(SON).Compared with DT,the NT and SS treatments improved soil available nutrients in the 0-20 cm layer.These findings suggest that SS is an excellent practice for increasing soil carbon,nitrogen and nutrient availability in dryland wheat fields in North China.

Effects of vacancy and external electric field on the electronic properties of the MoSi_(2)N_(4)/graphene heterostructure

Recently,the newly synthesized septuple-atomic layer two-dimensional(2D)material MoSi_(2)N_(4)(MSN)has attracted attention worldwide.Our work delves into the effect of vacancies and external electric fields on the electronic properties of the MSN/graphene(Gr)heterostructure using first-principles calculation.We find that four types of defective structures,N-in,N-out,Si and Mo vacancy defects of monolayer MSN and MSN/Gr heterostructure are stable in air.Moreover,vacancy defects can effectively modulate the charge transfer at the interface of the MSN/Gr heterostructure as well as the work function of the pristine monolayer MSN and MSN/Gr heterostructure.Finally,the application of an external electric field enables the dynamic switching between n-type and p-type Schottky contacts.Our work may offer the possibility of exceeding the capabilities of conventional Schottky diodes based on MSN/Gr heterostructures.