基于卷积神经网络和双向门控循环单元网络注意力机制的情感分析  

Sentiment Analysis of Attention Mechanism Based on Convolutional Neural Network and Bidirectional Gated Recurrent Unit Network

作　　者：张腾 刘新亮[1,2,3] 高彦平 ZHANG Teng;LIU Xin-liang;GAO Yan-ping(Department of Computer and Information Engineering,Beijing Technology and Business University,Beijing 100048,China;National Engineering Laboratory for Agri-product Quality Traceability,Beijing 100048,China;Department of Information and Electrical Engineering,China Agricultural University,Beijing 100083,China)

机构地区：[1]北京工商大学计算机与信息工程学院,北京100048 [2]农产品质量安全追溯技术及应用国家工程实验室,北京100048 [3]中国农业大学信息与电气工程学院,北京100083

出　　处：《科学技术与工程》2021年第1期269-274,共6页Science Technology and Engineering

基　　金：国家重点研发计划(2016YFD0401205);北京市科委科技计划(Z191100008619007);北京市教委科研计划(KM201510011008)。

摘　　要：传统的情感分析方法不能获取全局特征,以及否定词、转折词和程度副词的出现影响句子极性判断。在深度学习方法基础上提出了基于卷积神经网络和双向门控循环单元网络注意力机制的短文本情感分析方法。将情感积分引入卷积神经网络,利用情感词自身信息,通过双向门控循环网络模型获取全局特征,对影响句子极性的否定词、转折词和程度副词引入注意力机制实现对这类词的重点关注,提取影响句子极性的重要信息。实验结果表明,该模型与现有相关模型相比,有效提高情感分类的准确率。Sentiment analysis has been one of the research hotspots of natural language processing.The traditional sentiment analysis method cannot well obtain the global features,and negative words,turning words and degree adverbs affect the sentence polarity judgment.For a deep learning method,a short text sentiment analysis was proposed using convolutional neural networks(CNN)and the attention mechanism of the bidirectional gated recurrent unit(Bi-GRU).In view of the CNN,the emotional integration was introduced to extract the sentimental information of an emotional word,and the global feature was abstracted through the Bi-GRU.Furthermore,the negative words,turning words,and degree adverbs which affecting the polarity of the sentence are introduced into the attention mechanism while important information was retained.The experimental results show that the model can effectively improve the accuracy of sentiment classification better than those of the existing related ones.

关 键 词：深度学习 双向门控循环单元(Bi-GRU) 注意力机制 卷积神经网络 情感分析 

分 类 号：TP393.01[自动化与计算机技术—计算机应用技术]