基于RoBERTa和位置特征的多注意力方面级情感分析

面向方面的情感分析是一种细粒度的情感分析,目的是识别句子中特定方面的情感极性。现有模型中存在句子和方面语义丢失以及语义提取不充分的问题,并且忽略句子中的上下文信息和方面术语的位置关系,导致情感极性识别准确率下降。因此,提出了基于RoBERTa和位置特征的多注意力方面级情感分析模型用于面向方面级别的情感分类。为验证模型的有效性,在SemEval2014餐厅14、SemEval2014笔记本电脑14、SemEval2015餐厅15、SemEval2016餐厅16数据集中进行了大量的实验,结果表明所提出模型的准确率和F 1高于比较的基准模型。

基于位置信息多注意交互模型的方面级情感分析

方面级情感分析是用于识别文本中特定方面情感倾向的细粒度级情感分析。针对传统神经网络模型在上下文和方面词之间交互性较少,不同距离上下文对方面词情感极性判断有不同影响等问题,提出了基于位置信息多注意交互模型(MAIP)。该模型引入依存句法树表示词语之间的依存关系,将位置信息嵌入到词向量中,使用双向GRU获取方面词和上下文的隐藏向量表示,并利用多注意力机制学习词语之间的交互信息。实验结果表明,该模型能有效提升情感分析的准确度。

基于多注意力Faster RCNN的噪声干扰下印刷电路板缺陷检测

针对工业环境中噪声干扰导致的印刷电路板缺陷检测困难的问题,提出基于多注意力Faster RCNN的印刷电路板缺陷检测方法,分别于特征提取以及特征融合部分引入注意力机制以获得具有抗干扰能力的特征表示,提升检测效果。首先使用分离注意力网络提取缺陷特征,使网络自动关注到缺陷特征,以降低噪声干扰的影响;其次,使用平衡特征金字塔融合不同分辨率特征,利用非局部注意力机制对融合特征进行全局感受野内不同区域特征的加权,增强其缺陷表征能力并进一步抑制噪声干扰;最后,依据所获得的特征表示,利用区域建议网络生成缺陷候选框并利用全连接层对其进行位置以及类别的确定以得到检测结果。在不同程度噪声干扰下的印刷电路板缺陷数据集上进行实验验证,平均检测精度达到92.4%,证明了所提方法的有效性和可行性。

基于多注意力融合的抗遮挡目标跟踪

针对基于孪生网络的目标跟踪算法在相似目标干扰和发生遮挡时容易丢失目标的问题,提出一种基于多注意力融合的抗遮挡目标跟踪算法(anti-occlusion target tracking based on multi-attention fusion, AOTMAF)。为更好地模拟遮挡图片,引入渐进式随机遮挡模块,由易到难地随机生成遮挡块对图像进行多区域遮挡,通过人工模拟被遮挡图像的方式扩充负样本数据集,提升模型在遮挡情况下对判别性特征的提取能力。从深度、高度与宽度三个维度挖掘特征图通道信息,并通过融合空间注意力,聚合特征图上每个位置的空间依赖性,增强特征表达能力,进一步提高跟踪的鲁棒性。实验结果表明,在OTB100、VOT2018、GOT-10K公开数据集上,本研究方法在复杂场景下能有效提升跟踪精度和鲁棒性。

面向聊天机器人的多注意力记忆网络

如何对多轮的对话历史进行建模和推理是构建一个智能聊天机器人的主要挑战之一.基于循环或门控的记忆网络已经被证明是进行对话建模的有效方式.然而,这种方式有两个缺点,一是使用复杂的循环结构,导致计算效率较低;二是使用代价较大的强监督信息或先验信息,不利于扩展和迁移应用到新的领域.针对上述问题,本文提出了一种端到端的多注意力记忆网络.首先,该网络采取结合词向量和位置编码的方式对文本输入进行表示;其次,使用并行的多层注意力在不同子空间捕获对话交互中的关键信息来更好地建模对话历史;最后,通过捷径连接的方式叠加多注意力层管理信息流,实现对建模结果的多次推理.在bAbI-dialog数据集上的实验表明,该网络可以有效地对多轮对话进行建模和推理,而且具有较好的时间性能.

基于多注意力和变分编码时序网络的发动机剩余使用寿命预测方法

准确可靠的剩余使用寿命(RUL)预测结果可为决策者提供有价值的信息,以采取适当的维护策略,最大限度地利用设备,避免昂贵的故障维修费用;为了从高噪声的真实工况中对发动机故障进行有效诊断,提出了一种融合多注意力机制和变分编码的时序预测模型(MA-VBLSTM),通过嵌入多注意力机制获得所有特征在空间维度和通道维度的不同权重,以提高退化特征的提取能力;采用变分编码器进行退化信息编码并学习数据间深度隐藏的信息;利用双向长短时记忆网络的长短期时序数据双向处理能力实现发动机RUL的预测;实验结果表明,该模型在发动机CMAPSS数据集的FD001、FD002、FD003、FD004子数据集上,RMSE和Score值相比现有方法分别平均降低5.27%和10.70%、1.37%和1.68%、6.37%和26.94%、3.02%和2.06%。

基于神经网络的多注意属性情感分析

针对传统神经网络无法准确捕捉多个目标属性情感特征的问题,提出一种基于神经网络的多注意属性情感分析模型。用双向长短时记忆提取上下文和目标属性之间关联,用位置编码机制捕捉与目标属性相邻词的重要信息,结合位置编码和强注意力机制挖掘上下文和目标属性进行情感分析,在SemEval 2014和twitter数据集进行实验。实验结果表明,该模型在laptop、restaurant和twitter数据集上的准确率分别是72.45%、80.17%和73.39%,相比传统神经网络准确率得到了提高。

基于用电多注意力的独居老人风险识别方法

为解决中国城市独居老人健康监护难的问题,本文提出一种基于用电多注意力的独居老人风险识别方法。通过新一代智能电表采集独居老人的历史用电数据,提取独居老人各类型电器的用电特征;通过多注意力机制学习独居老人的电量趋势、负荷波动等特征,建立独居老人用电指纹。在此基础上,将当前独居老人的用电数据与用电指纹数据进行比对,判断独居老人是否存在风险。在某城市社区应用该方法,识别准确率达96.7%。该方法通过电力数据实现了独居老人风险识别,提升了政府的独居老人管理水平。

基于弱监督多注意融合网络的细粒度图像识别

针对细粒度图像识别任务中常见的判别性区域难以定位的问题,提出了一种弱监督多注意融合网络,该网络通过两种注意力模块的组合实现判别性区域的准确定位。其中,双域自注意力模块将多种注意力结合起来,强化模型对关键特征的提取。混合卷积注意力融合模块分别通过并行和串行架构融合不同尺度的注意力,充分获取特征间的全局及局部联系。实验结果表明,所提出的方法是有效的,与基线模型的结果相比有较大幅度的提升。

基于多注意力机制的红外与可见光图像夜间目标检测

目标检测一直是计算机视觉领域的研究热点,YOLO系列目标检测模型已广泛应用于多个领域。然而,目前关于目标检测的图像数据大多是基于单一类型传感器,难以完整地表征成像场景,且检测到的目标所包含有用信息具有局限性,尤其是在低照度、夜晚、雨雾等条件下,目标检测更加困难。为了更好地检测夜间目标,本文提出了一种结合CBAM注意力机制与Transformer的多注意力机制的红外与可见光图像夜间目标检测方法,通过添加Transformer来获取丰富的局部和上下文信息,通过添加CBAM注意力机制来减少误检。为了验证方法的有效性,本文选取了5种当前主流的目标检测算法在公开红外目标检测数据集上进行测试,本文方法与原始YOLO v7相比,m AP从62.6%提升至71.5%。本文还制作了一个用于夜间目标检测红外-可见光融合目标检测数据集。在该数据集上与原始YOLOv7相比,mAP从79.90%提升至94.80%,效果非常显著。

一种多粒度多注意力交互匹配网络模型在金融领域智能问答系统的对话改进研究

【目的】针对金融领域智能问答系统多轮对话中存在的回复不准确、返回多结果中精确度不足以及一词多义问题,构建多粒度多注意力交互匹配模型。【方法】提出基于BERT的多粒度多注意力交互匹配模型(MGMAI),MGMAI包含预处理层、表示层、注意力交互层、语义聚合层和对话选择层,聚焦于对话中的关键信息并利用这些信息实现高效的对话匹配。【结果】MGMAI模型被应用于两个公开的多轮对话数据集上进行训练与验证,并在金融领域多轮对话数据上进行微调,实验结果表明,MGMAI在R10@1、R10@2、R10@5指标上比先进的深度语境建模模型高出0.019、0.010和0.007。【局限】仅在金融领域智能问答系统中进行测试,未在其他领域或更多样化的数据集上验证模型的泛化能力。【结论】MGMAI模型能有效提升金融领域智能问答系统中的多轮对话准确度和系统处理多义性问题的能力,对于类似的多轮对话系统,该模型展现出潜在的应用价值和改进空间。

复杂场景下多尺度船舶实时检测方法

船舶检测在军事侦察、海上目标跟踪、海上交通管制等任务中发挥着重要作用。然而,受船舶外形尺度多变和复杂海面背景的影响,在复杂海面上检测多尺度船舶仍然是一个挑战。针对此难题,提出了一种基于多层信息交互融合和注意力机制的YOLOv4改进方法。该方法主要通过多层信息交互融合(multi-layer information interactive fusion,MLIF)模块和多注意感受野(multi-attention receptive field,MARF)模块构建一个双向细粒度特征金字塔。其中,MLIF模块用于融合不同尺度的特征,不仅能将深层的高级语义特征串联在一起,而且将较浅层的丰富特征进行重塑;MARF由感受野模块(receptive field block,RFB)与注意力机制模块组成,能有效地强调重要特征并抑制冗余特征。此外,为了进一步评估提出方法的性能,在新加坡海事数据集(Singapore maritime dataset,SMD)上进行了实验。实验结果表明,所提方法能有效地解决复杂海洋环境下多尺度船舶检测的难题,且同时满足了实时需求。

Improved YOLOv5-Based Inland River Floating Garbage Detection Model

Detection of floating garbage in inland rivers is crucial for water environmental protection,as it effectively reduces ecological damage and ensures the safety of water resources.To address the inefficiency of traditional cleanup methods and the challenges in detecting small targets,an improved YOLOv5 object detection model was proposed in this study.In order to enhance the model’s sensitivity to small targets and mitigate the impact of redundant information on detection performance,a bi-level routing attention mechanism was introduced and embedded into the backbone network.Additionally,a multi-scale detection head was incorporated into the model,allowing for more comprehensive coverage of floating garbage of various sizes through multi-scale feature extraction and detection.The Focal-EIoU loss function was also employed to optimize the model parameters,improving localization accuracy.Experimental results on the publicly available FloW_Img dataset demonstrated that the improved YOLOv5 model outperforms the original YOLOv5 model in terms of precision and recall,achieving a mAP(mean average precision)of 86.12%,with significant improvements and faster convergence.

Improved YOLOv8-Based Target Detection Algorithm for UAV Aerial Image

In response to the challenge of low detection accuracy and susceptibility to missed and false detections of small targets in unmanned aerial vehicles(UAVs)aerial images,an improved UAV image target detection algorithm based on YOLOv8 was proposed in this study.To begin with,the CoordAtt attention mechanism was employed to enhance the feature extraction capability of the backbone network,thereby reducing interference from backgrounds.Additionally,the BiFPN feature fusion network with an added small object detection layer was used to enhance the model's ability to perceive for small objects.Furthermore,a multi-level fusion module was designed and proposed to effectively integrate shallow and deep information.The use of an enhanced MPDIoU loss function further improved detection performance.The experimental results based on the publicly available VisDrone2019 dataset showed that the improved model outperformed the YOLOv8 baseline model,mAP@0.5 improved by 20%,and the improved method improved the detection accuracy of the model for small targets.

基于知识图谱的问答系统中属性映射方法研究

在基于知识图谱的智能问答系统中,属性映射模块结果的错误传播会导致最终无法得到正确答案,对此提出了一种基于多注意力多维文本的属性映射方法。首先通过拆分问题文本及结合属性信息得到多维文本表示;其次使用长短期记忆网络(long-short-term memory,LSTM)层生成各自的隐层表示;然后输入多注意力机制层后使问句和属性之间的关系及语义信息更加完善,利用属性之间的交互信息及多种角度来加强问句语义信息的理解;最后通过卷积神经网络(convolutional neural networks,CNN)提取局部特征并且采用softmax分类器实现属性映射。试验结果表明,在自然语言处理与中文计算会议(NLPCC 2018)中知识库问答(KBQA)任务所提供的开源数据集上,本方法相比主流属性映射模型其性能有显著提升,准确率最高提升6.62%。本模型可以补足单一文本表示与注意力机制的短板,有效解决属性映射模块中语义歧义的问题,这有助于后续提高智能问答系统的整体性能。

冬小麦杂草防治要点

本文主要介绍了冬小麦使用除草剂时的时机、选用除草剂配方的方法、不同配方除草方向及喷打除草剂时应该注意的技术问题,给冬小麦化学除草时提供技术参考。

变工况的滚动轴承智能故障诊断方法

针对变工况下滚动轴承故障关键特征难提取的问题,提出了基于多尺度多注意力融合的一维卷积神经网络滚动轴承故障诊断方法。采用多尺度模块提取1DCNN对多尺度特征,通过多注意力机制加强了1DCNN对关键信息提取能力,将多尺度模块和多注意力机制融合,给出了多尺度和多注意力融合方法,构建了多尺度多注意力融合的1DCNN。以美国凯斯西储大学开放轴承数据库样本为测试对象,在不同噪声和不同载荷的工况下,将该方法与其他深度学习方法进行了对比分析。结果表明,该方法在不同载荷和不同噪声条件下平均准确率在99%以上,在T-SNE可视化下的分类效果也更显著。

A multi-attention RNN-based relation linking approach for question answering over knowledge base

Aiming at the relation linking task for question answering over knowledge base,especially the multi relation linking task for complex questions,a relation linking approach based on the multi-attention recurrent neural network(RNN)model is proposed,which works for both simple and complex questions.First,the vector representations of questions are learned by the bidirectional long short-term memory(Bi-LSTM)model at the word and character levels,and named entities in questions are labeled by the conditional random field(CRF)model.Candidate entities are generated based on a dictionary,the disambiguation of candidate entities is realized based on predefined rules,and named entities mentioned in questions are linked to entities in knowledge base.Next,questions are classified into simple or complex questions by the machine learning method.Starting from the identified entities,for simple questions,one-hop relations are collected in the knowledge base as candidate relations;for complex questions,two-hop relations are collected as candidates.Finally,the multi-attention Bi-LSTM model is used to encode questions and candidate relations,compare their similarity,and return the candidate relation with the highest similarity as the result of relation linking.It is worth noting that the Bi-LSTM model with one attentions is adopted for simple questions,and the Bi-LSTM model with two attentions is adopted for complex questions.The experimental results show that,based on the effective entity linking method,the Bi-LSTM model with the attention mechanism improves the relation linking effectiveness of both simple and complex questions,which outperforms the existing relation linking methods based on graph algorithm or linguistics understanding.

Multi-head attention-based long short-term memory model for speech emotion recognition

To fully make use of information from different representation subspaces,a multi-head attention-based long short-term memory(LSTM)model is proposed in this study for speech emotion recognition(SER).The proposed model uses frame-level features and takes the temporal information of emotion speech as the input of the LSTM layer.Here,a multi-head time-dimension attention(MHTA)layer was employed to linearly project the output of the LSTM layer into different subspaces for the reduced-dimension context vectors.To provide relative vital information from other dimensions,the output of MHTA,the output of feature-dimension attention,and the last time-step output of LSTM were utilized to form multiple context vectors as the input of the fully connected layer.To improve the performance of multiple vectors,feature-dimension attention was employed for the all-time output of the first LSTM layer.The proposed model was evaluated on the eNTERFACE and GEMEP corpora,respectively.The results indicate that the proposed model outperforms LSTM by 14.6%and 10.5%for eNTERFACE and GEMEP,respectively,proving the effectiveness of the proposed model in SER tasks.

Adaptive multi-modal feature fusion for far and hard object detection

In order to solve difficult detection of far and hard objects due to the sparseness and insufficient semantic information of LiDAR point cloud,a 3D object detection network with multi-modal data adaptive fusion is proposed,which makes use of multi-neighborhood information of voxel and image information.Firstly,design an improved ResNet that maintains the structure information of far and hard objects in low-resolution feature maps,which is more suitable for detection task.Meanwhile,semantema of each image feature map is enhanced by semantic information from all subsequent feature maps.Secondly,extract multi-neighborhood context information with different receptive field sizes to make up for the defect of sparseness of point cloud which improves the ability of voxel features to represent the spatial structure and semantic information of objects.Finally,propose a multi-modal feature adaptive fusion strategy which uses learnable weights to express the contribution of different modal features to the detection task,and voxel attention further enhances the fused feature expression of effective target objects.The experimental results on the KITTI benchmark show that this method outperforms VoxelNet with remarkable margins,i.e.increasing the AP by 8.78%and 5.49%on medium and hard difficulty levels.Meanwhile,our method achieves greater detection performance compared with many mainstream multi-modal methods,i.e.outperforming the AP by 1%compared with that of MVX-Net on medium and hard difficulty levels.