摘要
针对传统卷积神经网络只提取局部短语特征而忽略了上下文的句子特征,影响了文本分类效果这一问题,提出一种基于BiGRU网络和胶囊网络的文本情感分析模型。采用联合神经网络,利用Glove模型预训练词向量,将其输入到双向门限循环单元(BiGRU)模型进行序列化学习得到上下文特征;添加胶囊网络(capsule network)模型,提取深层次短语特征;交给分类器进行情感分类。通过在IMDB数据集上进行实验,验证该方法有效提高了文本分类的准确率。
Aiming at the problem that the traditional convolutional neural network only extracts local phrase features and ignores contextual sentence features,which affects the effect of text classification,a text sentiment analysis model based on BiGRU network and capsule network was proposed.A joint neural network was used.The Glove model was used to pre-train the word vector and it was inputted into the bi-directional threshold recurrent unit(BiGRU)model for serialization learning to obtain contextual features.A capsule network model was added to extract deep-level phrases features that were later handed over to the classifier for sentiment classification.Experiments on IMDB data set verify that the proposed method effectively improves the accuracy of text classification.
作者
李多娇
何成万
雷力
LI Duo-jiao;HE Cheng-wan;LEI Li(College of Computer Science and Engineering,Wuhan Institute of Technology,Wuhan 430205,China)
出处
《计算机工程与设计》
北大核心
2022年第2期540-545,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61272115)
武汉工程大学教育创新基金项目(CX2019235、CX2019236、CX2019237)。