[1]胡春涛,夏玲玲*,张 亮,等.基于胶囊网络和卷积网络的文本分类对比[J].计算机技术与发展,2020,30(10):86-91.[doi:10. 3969 / j. issn. 1673-629X. 2020. 10. 016]
　HU Chun-tao,XIA Ling-ling *,ZHANG Liang,et al.Comparative Study of News Text Classification Based on Capsule Network and Convolution Network[J].,2020,30(10):86-91.[doi:10. 3969 / j. issn. 1673-629X. 2020. 10. 016]
点击复制

基于胶囊网络和卷积网络的文本分类对比()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
30
期数:
2020年10期
页码:
86-91
栏目:
智能、算法、系统工程
出版日期:
2020-10-10

文章信息/Info

Title:
Comparative Study of News Text Classification Based on Capsule Network and Convolution Network
文章编号:
1673-629X(2020)10-0086-06
作者:
胡春涛1; 夏玲玲1*; 张 亮1; 王 超2; 韩 旭1
1. 江苏警官学院 计算机信息与网络安全系,江苏 南京 210031; 2. 南京市玄武分局,江苏 南京 210000
Author(s):
HU Chun-tao1; XIA Ling-ling1 *; ZHANG Liang1; WANG Chao2; HAN Xu1
1. Department of Computer Information and Cyber Security,Jiangsu Police Institute,Nanjing 210031,China; 2. Nanjing Police Station Xuanwu Sub-bureau,Nanjing 210000,China
关键词:
文本分类; 神经网络; 胶囊网络; 卷积网络; 网络舆情
Keywords:
text classification; neural network; capsule network; CNN; network public opinion
分类号:
TP391. 1
DOI:
10. 3969 / j. issn. 1673-629X. 2020. 10. 016
摘要:
针对 CNN 应用于文本分类任务中,存在固定大小的卷积核会限制复杂语言结构特征的提取,以及 CNN 在最大值池化时会丢失大量特征信息这两类问题,引入胶囊网络替代 CNN 提取文本特征,根据 Sabour 等人提出的动态路由算法作为胶囊网络内部的传播原理,并将其与 LSTM 连接形成融合神经网络模型,并与文本分类任务中常用的 CNN,LSTM,LSTM 连接 CNN 三个基线模型进行比较。实验结果表明,胶囊网络相较于使用卷积核的 CNN 能够不受特征检测器结构和大小的限制,更加灵活地学习文本整体与部分的内在空间关系;不仅能够检测固定的文本特征,而且能够检测特征的变体及其可能性。LSTM 连接胶囊网络的模型能有效提升文本分类的 F1值,达到 0. 932,相比其他基线模型拥有最佳性能。
Abstract:
Concerning the problems that while dealing with text classification tasks,the fixed-volume convolution kernel of CNN will limit the extraction of comp-lex language structure features,and max pooling operation will lose a lot of feature information at the pooling layer,we use an improved fusion model of LSTM connected capsule network which uses the dynamic routing algorithm proposed by Sabour et al. as the internal propagation principle instead of CNN for text features extracting. The experiments are conducted to investigate the performance of improved model by comparing with other three baseline models such as CNN, LSTM and LSTM connected CNN. It is found that the capsule network can learn the inherent spatial relationship between the whole and the part of the context more flexibly than the CNN using the convolution kernel. It can not only detect the fixed text features,and can detect variants of features and their possibilities. Te improved model of LSTM connected capsule network can effectively improve the F1 value to 0.932,which has the best performance compared with other baseline models.

相似文献/References:

[1]田昕辉 李成基.带有短语切分的中文文本分类方法[J].计算机技术与发展,2010,(01):5.
　TIAN Xin-hui,LEE Sung-kee.Phrase Segmentation for Chinese Text Classification[J].,2010,(10):5.
[2]姜鹤 陈丽亚.SVM文本分类中一种新的特征提取方法[J].计算机技术与发展,2010,(03):17.
　JIANG He,CHEN Li-ya.A New Feature Selection Method in SVM Text Categorization[J].,2010,(10):17.
[3]路川 胡欣杰.区域航空市场航线客流量预测研究[J].计算机技术与发展,2010,(04):84.
　LU Chuan,HU Xin-jie.Analysis of Regional Airline Passenger Forecast Title[J].,2010,(10):84.
[4]周瑛 张铃.有限混合模型在文本分类中的应用研究[J].计算机技术与发展,2010,(06):18.
　ZHOU Ying,ZHANG Ling.Study of Application of Finite Mixture Model in Text Classification[J].,2010,(10):18.
[5]高峥 陈蜀宇 李国勇.混合入侵检测系统的研究[J].计算机技术与发展,2010,(06):148.
　GAO Zheng,CHEN Shu-yu,LI Guo-yong.Research of a Hybrid Intrusion Detection System[J].,2010,(10):148.
[6]许幸 张启蕊.基于KNN算法的医药信息文本分类系统的研究[J].计算机技术与发展,2009,(04):206.
　XU Xing,ZHANG Qi-rui.Research of Medical Information Text Categorization Based on KNN Algorithm[J].,2009,(10):206.
[7]包力伟 周俊.铸锻企业生产质量控制系统的开发[J].计算机技术与发展,2008,(04):174.
　BAO Li-wei,ZHOU Jun.Development of a Manufacture Quality Control System in Casting Company[J].,2008,(10):174.
[8]李志俊 程家兴 金奎 饶玉佳.基于样本期望训练数的BP神经网络改进研究[J].计算机技术与发展,2009,(05):103.
　LI Zhi-jun,CHENG Jia-xing,JIN Kui,et al.BP Algorithm Improvement Based on Sample Expected Training Number[J].,2009,(10):103.
[9]李龙澍 葛瑞峰 王慧萍.基于神经网络的批强化学习在Robocup中的应用[J].计算机技术与发展,2009,(07):98.
　LI Long-shu,GE Rui-feng,WANG Hui-ping.Application of Batch Reinforcement Learning Based on NN to Robocup[J].,2009,(10):98.
[10]陈锦禾 范新 沈闻 沈洁.基于情感词识别的BBS情感分类研究[J].计算机技术与发展,2009,(07):120.
　CHEN Jin-he,FAN Xin,SHEN Wen,et al.Research on Sentiment Classification of BBS Reviews Based on Identifying Words with Polarity[J].,2009,(10):120.
[11]孟杰 耿正 严莉莉 张燕平.覆盖算法在文本分类中的应用[J].计算机技术与发展,2007,(07):183.
　MENG Jie,GENG Zheng,YAN Li-li,et al.Application of Cover Algorithm in Text Categorization[J].,2007,(10):183.
[12]李丰翼,刘万里,杨晓辉,等.基于多头自注意力机制与 CNN 的文本分类模型[J].计算机技术与发展,2022,32(S1):18.[doi:10. 3969 / j. issn. 1673-629X. 2022. S1. 004]
　LI Feng-yi,LIU Wan-li,YANG Xiao-hui,et al.Text Classification Model Based on Multi-headed Self-attention Mechanism and CNN[J].,2022,32(10):18.[doi:10. 3969 / j. issn. 1673-629X. 2022. S1. 004]

更新日期/Last Update: 2020-10-10