摘要
目前,深度学习在文本识别方面已经达到了相当高的准确率,但是在文本的情感识别方面还未达到理想的效果。针对传统卷积神经网络在词向量构建和卷积池化部分的一些不足,提出了一种新的情感模型——基于分段多池卷积神经网络(piecewise multi-pooling convolution neural network,PMPCNN)模型。该模型分别从情感词向量的构造、卷积层、池化层和应用Dropout算法防止模型过拟合等多方面入手进行改进。大量的对比试验数据表明,相比传统卷积神经网络,该模型具有更为良好的实验效果和准确率。
At present,deep learning has achieved a high degree of accuracy in text recognition,but it has not achieved the de⁃sired effect in the emotional recognition of text.Aiming at some shortcomings of traditional convolutional neural networks in word vector construction and convolution pooling,this paper proposes a new emotion model based on piecewise multi-pooling convolution neural network(PMPCNN)model.The model is improved from the construction of the emotion word vector,the convolution layer,the pooling layer and the application of the Dropout algorithm to prevent the model from overfitting.A large number of comparative experimental data show that the model has better experimental results and accuracy than traditional convolutional neural networks.
作者
付晓杰
张曦煌
FU Xiaojie;ZHANG Xihuang(School of Internet of Things Engineering,Jiangnan University,Wuxi 214000)
出处
《计算机与数字工程》
2020年第11期2665-2670,共6页
Computer & Digital Engineering
基金
江苏省产学研合作项目基金(编号:BY2015019-30)资助。
关键词
情感识别
卷积神经网络
情感词向量
Dropout算法
emotional recognition
convolutional neural network
emotion word vector
Dropout algorithm
作者简介
付晓杰,男,硕士研究生,研究方向:卷积神经网络的应用与研究;张曦煌,男,博士,教授,研究方向:计算机应用技术,计算机信息管理,嵌入式系统与结构。
