摘要
结合卷积神经网络CNN和最小门控单元MGU各自的优势,融合注意力机制,提出注意力C_MGU神经网络模型。通过CNN的卷积层模块捕捉提取文本的初步特征表示,利用Attention机制和MGU模块对文本的初步特征表示进行关键信息的加强和优化,并将生成的文本深层特征表示输入到Softmax层进行回归处理。对公开数据集IMBD、Sentiment140进行情感分类实验,结果表明该模型能够强化对文本的句义理解,可进一步学习序列相关特征,有效地提高情感分类的准确率。
Combining the advantages of convolutional neural network(CNN)and minimum gating unit(MGU),and integrating the attention mechanism,this paper proposes an attention C_MGU neural network model.Through the CNN convolutional layer module,the preliminary feature representation of the extracted text was captured,and the key information of the preliminary feature representation of text was further enhanced and optimized by using the attention mechanism and the MGU module.The generated text deep feature representation was input into the Softmax layer for regression processing.Sentiment classification experiments are carried out on the open data sets IMBD and sentiment140.The results show that our model can enhance the understanding of sentence meaning,further learn the sequence related features,and effectively improve the accuracy of sentiment classification.
作者
徐菲菲
芦霄鹏
Xu Feifei;Lu Xiaopeng(School of Computer Science and Technology,Shanghai Electric Power University,Shanghai 200090,China)
出处
《计算机应用与软件》
北大核心
2020年第9期75-80,125,共7页
Computer Applications and Software
基金
国家自然科学基金项目(61305094)。
