摘要
为了解决传统的文本极性智能判断方法判断结果准确率和召回率普遍较低的问题,基于改进深度学习算法研究一种新的文本极性智能判断方法。在CNN结构基础上设计一种新的深度学习算法模型,模型由输入层、输出层、采集层、连接层、卷积层五部分构成。使用该模型对文本进行智能判断,判断过程共有五步,分别是文本预处理、情感词提取、表情符号提取、感情倾向值计算和情感最终倾向值分析。为检测所提方法的有效性以及优越性,与传统判断方法进行实验对比,结果表明,基于改进深度学习算法的文本极性智能判断方法判断的准确率和召回率更高,发展空间更广阔。
The accuracy and recall rate of traditional text polarity intelligent judgment methods both are generally low.In view of the above,a new method of text polarity intelligent judgment is studied based on improved deep learning algorithm.A new deep learning algorithm model is designed based on the CNN structure.The model consists of five parts:input layer,output layer,acquisition layer,connection layer and convolution layer.This model is used for text intelligent judgment.The judgment process is devided into five steps:text preprocessing,emotion word extraction,expression symbol extraction,emotion tendency value calculation and emotion final tendency value analysis.In order to test the effectiveness and superiority of the proposed method,an experimental comparison with the traditional judgment method was performed.The results show that the judgemental accuracy and recall rate of the text polarity intelligent judgment method based on the improved deep learning algorithm is higher,and the development space is broader.
作者
宋思晗
王兴芬
杜惠英
SONG Sihan;WANG Xingfen;DU Huiying(School of Information Management,Beijing Information Science&Technology University,Beijing 100192,China)
出处
《现代电子技术》
北大核心
2020年第1期76-79,85,共5页
Modern Electronics Technique
基金
国家自然科学基金面上项目：网络零售交易风险动态评估及预警研究（71571021）
关键词
文本极性
智能判断方法
算法模型设计
有效性检测
深度学习算法
文本预处理
text polarity
intelligent judgment method
algorithm model design
effectiveness detection
deep learning algorithm
text pre⁃processing
