基于CNN-BiLSTM网络引入注意力模型的文本情感分析  

Text Sentiment Analysis Based on CNN-BiLSTM Network and Attention Model

作　　者：王丽亚 刘昌辉[1] 蔡敦波[1] 赵彤洲[1] 王梦 WANG Liya;LIU Changhui;CAI Dunbo;ZHAO Tongzhou;WANG Meng(School of Computer Science and Engineering,Wuhan Institute of Technology,Wuhan 430205,China)

机构地区：[1]武汉工程大学计算机科学与工程学院

出　　处：《武汉工程大学学报》2019年第4期386-391,共6页Journal of Wuhan Institute of Technology

基　　金：国家自然科学基金(61103136);武汉工程大学研究生教育创新计划项目(CX2018196)

摘　　要：为了解决单一卷积神经网络(CNN)缺乏利用文本上下文信息的能力和简单循环神经网络(RNN)无法解决长时依赖的问题,提出CNN-BiLSTM网络引入注意力模型的文本情感分析方法。首先利用CNN的特征强学习能力提取局部特征,再利用双向长短时记忆网络(BiLSTM)提取上下文相关特征的能力进行深度学习,最后,增加注意力层获取重要特征,使模型提取到有效的特征。在IMDB数据集上Accuracy值和均方根误差(RMSE)值分别达到90.34%和0.296 7,在Twitter数据集上Accuracy值和RMSE值分别达到76.90%、0.417 4,且模型时间代价小。结果表明,本文提出的模型有效提升了文本分类的准确率。To solve the problems that single Convolutional Neural Network(CNN)lacks the ability to utilize text context information and simple Recurrent Neural Network(RNN)cannot deal with long-term dependence,we propose a text sentiment analysis method by introducing the attention model into a CNN-BiLSTM network. In the CNN-BiLSTM network,CNN model and Bidirectional Long Short-Term Memory(BiLSTM)model are used to extract local features and context-related features,respectively. After that,the attention layer is used to focus our attention on the most critical features. Experiments were performed on both IMDB and Twitter datasets. The accuracy and Root Mean Squared Error(RMSE) achieved on IMDB dataset are 90.34% and 0.296 7,respectively. As to dataset Twitter,76.90% accuracy and 0.417 4 RMSE are achieved. The experimental results show that our technique is able to improve the accuracy of text classification effectively with little runtime overhead.

关 键 词：卷积神经网络 CNN-BiLSTM 注意力机制 文本情感分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]