共查询到18条相似文献，搜索用时 46 毫秒
1.
针对基于主题的文本分类任务存在的主题特征表征能力不足、数据高维导致的特征维度高等问题,本文对输入的特征表示与卷积神经网络结构（CNN）做出了改进。在特征表示时提出了使用LDA模型计算逆主题空间频率从而得到文本的主题向量矩阵,降低了噪声主题的特征表达,增强了关键主题的权重;分别将文本的主题向量矩阵与词向量矩阵作为CNN模型的输入。提出了双层CNN网络结构,在每层CNN的池化层后增加一层多通道池化层,以融合每层CNN的池化结果,降低特征维度的同时获取更多的局部显著特征;最后使用Attention机制对融合的特征进行加权后输入到全连接层进行分类。由实验结果可知,改进的模型在文本分类任务上的准确率、召回率均在98%以上,F1值较基准实验提高了近6%。 相似文献
2.
电力调度系统在电力系统故障过程中会收到大量告警信号,若调度员无法在短时间内做出决策,则可能使故障扩大,为此提出基于告警信号文本挖掘的电力调度故障诊断方法,该方法包括告警信号文本预处理和故障诊断2个阶段。在第一阶段,基于隐马尔可夫模型(HMM)对告警信号文本进行分词并去除其中的停用词以构建本体词典,并采用向量空间模型(VSM)使文本向量化;在第二阶段,使用滑动时间窗读取实时告警信号,提出一种2层算法,第一层采用支持向量机(SVM)对滑窗内的告警信号进行分类,若分类结果判断为发生故障,则启动第二层k-均值聚类法提取较高可能性的故障供调度员参考。以某电力调度系统实际告警信号作为算例,验证了所提方法的可行性。 相似文献
3.
进度控制是水电工程管理的重要任务,及时总结进度管理信息有助于工程进度计划的制定与调整。水电工程建设中的进度信息多以半结构化、非结构化的文本形式呈现,增加了信息提取难度,实现水电工程进度文本信息自动化与智能化挖掘是当前亟待解决的问题。本文提出基于改进LDA的水电工程进度信息智能提取方法,智能提取进度管理文本中的关键信息。该方法基于传统LDA模型针对吉布斯采样机制,充分考虑词语间的关联关系,将原有随机单个采样过程改进为以共现度为基准的词对采样,强化了词语间的语义关联,提高了主题词语间的紧密性以及主题词语对主题描述的准确性。将所提出的方法应用于实际水电工程,对221份水电工程施工监理周报进行分析,共提取12个主题的工序关键词,并依照计算结果提取出主副工序;结果表明,改进LDA主题模型在水电工程进度文本工序特征词提取效果优于传统LDA主题模型,有助于提高工程施工进度关键工序词提取与信息挖掘效率,为水电工程施工智能化管理提供了新的手段。
4.
针对物联网中的评论等短文本进行情感分析时,出现上下文依赖性差和严重的特征稀疏,以及评论类文本的情感分析具有时效性等问题,提出了基于词嵌入和时间加权的高斯LDA算法(TG-LDA)。实验结果证明,与同类的主题模型相比,该模型的关键词的区分度强,主题的一致性高。 相似文献
5.
6.
短期光伏发电功率预测对维护电网安全稳定和协调资源利用具有重要的意义。提出了一种基于K均值算法(Kmeans)和支持向量机(SVM)的短期光伏发电功率预测方法。根据短期光伏发电特性和光伏发电季节特性,组织预测模型的训练样本集。通过K均值算法对训练样本集进行聚类分析,在聚类得到的各类别数据上分别训练支持向量机。预测时根据预测样本的类别使用相应的支持向量机进行发电功率预测。经实验表明所提出的方法相较于传统的BP、SVM模型精度有了明显的提升,具有较好的工程应用潜力。 相似文献
7.
8.
继电保护装置缺陷文本缺乏基于专业词典的数据挖掘,对继电保护缺陷定级、诊断和消除支撑不足,无法满足高效运维需求。结合某区域电网继电保护缺陷数据,提出了适用于继电保护装置缺陷的专业词典构建方法,并构建了相关专业词典。首先,汇总了该区域继电保护装置缺陷文本数据,形成缺陷文本语料库;其次,应用基于正则表达式的停用词识别方法,实现缺陷文本中无关字词的剔除;然后,采用机器与人工相结合的方法,构建了缺陷文本分词词典,采用潜在语义分析和决策树分类,实现了同义词合并;然后,通过整合停用词表、分词词典、同义词表,构建了该区域电网保护装置缺陷专业词典;最后,对比了使用词典前后的专业词汇齐普夫分布和语料库信息熵,验证了所构建专业词典的有效性。 相似文献
9.
由于船舶工业领域中的新闻内容篇幅较长且专业性较强,同时包含大量船舶领域专业词汇,目前针对该领域新闻文本分类的研究较少且缺少相应的船舶工业新闻语料.构建了一个船舶工业新闻语料库,并提出了一种新的面向船舶工业新闻的文本分类算法,首先基于文档频率、卡方统计量及主题模型LSA进行特征选择和特征降维,将文档-词矩阵映射成文档-主... 相似文献
10.
11.
抽取最佳鉴别特征是说话人辨认中的重要一步.本文在使用美尔倒谱系数(MFCC)及一阶差分组成的特征参数的基础上利用主分量分析(PCA)和线性判决分析(LDA)结合的提取方法,构造了一种新的特征参数.这种新的参数具有最佳鉴别特性,然后用支持向量机(SVM)对提取的特征分类辨认.实验结果表明该方法能更好地识别说话人,有更好的识别能力. 相似文献
12.
《IEEJ Transactions on Electrical and Electronic Engineering》2018,13(10):1446-1453
In this paper, a sentence‐level sentiment analysis method is proposed to deal with sentiment measurement and classification problems. It is developed from a model called the synthetic and computational language model (SCLM), which represents modifying and modified information, respectively, using matrices and vectors. In the proposed method, a global modifying matrix of a sentence is constructed, the determinant value of this matrix is calculated and adjusted, and then the final value is used as the sentiment value of the sentence. Regression experiment shows that the deviation between the output sentiment and the target sentiment does not exceed a class distance of five classes. The classification experiment shows that the proposed method has improved most of the performance compared to the simplified SCLM and in some cases, such as in ‘very positive’ class and ‘very negative’ class, reaches higher precision performance than the baseline method. © 2018 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc. 相似文献
13.
Xin Kang Fuji Ren Yunong Wu 《IEEJ Transactions on Electrical and Electronic Engineering》2016,11(6):768-775
Learning emotions from texts has been an active research topic in affective computing. However, the lack of reliable connection between emotions and language features has caused severely biased emotion predictions. Moreover, the author‐specific patterns in emotion expression could potentially affect emotion predictions, which has never been studied. In this paper, we propose a semisupervised learning algorithm to learn emotional features from large‐scaled micro‐blog documents with a Bayesian network, and introduce an emotion transition factor to generate the author‐specific emotion predictions. We infer the author‐specific emotions in micro‐blog streams through belief propagation, and learn the emotional features through an expectation maximization estimation procedure. We report results of single‐label and multilabel emotion predictions on a micro‐blog stream corpus, and analyze the improvements achieved by the semisupervised feature learning strategy and the incorporation of emotion transition patterns. Finally, we perform personality analysis based on the authors' emotion distribution and examine emotion distributions in the learned features. © 2016 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc. 相似文献
14.
15.
Enkhzol Dovdon Suvdaa Batsuuri 《IEEJ Transactions on Electrical and Electronic Engineering》2021,16(6):852-860
We introduce a novel approach for generating 2D RGB color images with a plot from the micro text (tweet) to be used for the overall polarity classification process of sentiment analysis. Researchers generally use word embedding and external resource-based embedding techniques for text preprocessing of sentiment analysis through machine learning, neural networks, and natural language processing approaches. We sought to identify alternative ways to represent tweets for text classification. According to the experimental results, using the new ‘Text2Plot’ representation method could increase F1 scores by 27.2% for Convolutional neural networks (CNNs), 10.3% for support vector machine, and 4.4% for random forest models compared to using simple vectors as features for sentiment analysis. Hence, we propose this new method as a useful text representation approach for sentiment analysis, natural language processing tasks, and image processing problems. © 2021 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC. 相似文献
16.
17.
为满足对金融市场的进一步了解以及股价预测的需求,结合投资者的情感倾向提出了一种基于TA/SA(technical analysis/sentimental analysis)的股票价格预测模型,建立投资者情感与未来股票价格之间的关系的方案.该方案主要包括获取情感指数,建立回归模型以及计算未来股票收盘价.利用该模型预测200只股票价格并与SVM和BP神经网络两种模型预测结果进行比较,结果显示所提出模型的预测正确率分别提高了1o.9％和7.4％,表明该模型具有更好的预测准确性和实用价值. 相似文献
18.
Yunong Wu Kenji Kita Kazuyuki Matsumoto 《IEEJ Transactions on Electrical and Electronic Engineering》2014,9(6):642-649
Emotion prediction has been a core task in affective computing, which aims at finding the thorough human mental states by analyzing people's activities. In this paper, we focus on predicting emotions in the public online blogs from different people, by extracting as many reasonable emotions for each blog sentence as possible. Concretely, we consider three different perspectives for analyzing the multiple emotions in a sentence: (i) predict sentence emotions by examining the emotion related topics in a global sense; (ii) predict the sentence emotions from the context‐sensitive word emotions; and (iii) predict sentence emotions by considering the emotional significance in the local bag of words. We build different probabilistic models from each perspective, to separately generate the sentence emotion probabilities. We then integrate these probabilistic models to jointly predict the emotion probabilities. Because the component models are based on different emotional assumptions with distinct features, the integrated predictions should predict emotions from more general perspectives and therefore yield better results. In the experiment, we employ different evaluation criteria to compare the multi‐emotion predictions from the single and integrated models. Compared to the results in the baseline model, our bi‐integrated models achieve 8.69% higher Micro F1 and 7.78% higher Macro F1 scores, on average. Moreover, our tri‐integrated model acquires 10.00% higher Micro F1 and 9.19% higher Macro F1 scores than the baseline results, which proves our assumption, and suggests interesting features in the different emotion perspectives. © 2014 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc. 相似文献