基于区域BLSTM和CNN的情感分析模型

Sentiment analysis based on regional BLSTM and CNN

结合传统的深度学习结构与注意力机制,提出基于区域双向长短期记忆网络与卷积神经网络相结合的情感分析模型。区域双向长短期记忆网络由3个互相关联的BLSTM构成,实现对目标词的重点关注,能更好提取全局信息;通过卷积神经网络获取局部特征,利用注意力池化比较局部特征与全局信息,保留句子中的重要信息。实验结果表明,提出的改进模型与现有相关模型相比取得了更好的分类效果。

Combining traditional deep learning structures with attention mechanism,a sentiment analysis model based on the combination of regional bidirectional long short term memory networks and convolutional neural networks was presented.The regional BLSTM was composed of three interrelated BLSTMs,which concentrated on target words,it extracted global information correctly.Local features were captured through convolutional neural networks simultaneously,and attention pooling was used to compare local features and global features,the model kept the important information in sentence.Experimental results show that the proposed model has better classification effects than the existing correlation model.