基于A-BiLSTM和CNN的文本分类
Text classification based on A-BiLSTM and CNN
黄远 1戴晓红 1黄伟建 1于钧豪 1黄峥1
作者信息
- 1. 河北工程大学信息与电气工程学院,河北邯郸 056038
- 折叠
摘要
为解决单一神经网络不能获取准确全局文本信息的问题,提出一种基于A-BiLSTM双通道和优化CNN的文本分类模型.A-BiLSTM双通道层使用注意力机制关注对文本分类贡献值较大的部分,并用BiLSTM提取文本中上下文语义信息;A-BiLSTM双通道层中将两者输出的特征信息融合,得到高级语义;A-BiLSTM双通道层后,使用优化CNN的强学习能力提取关键局部特征,得到最终文本特征表示.分类器输出文本信息的类别.实验结果表明,该模型分类效果优于其它对比模型,具有良好的泛化能力.
Abstract
To solve the problem that a single neural network cannot obtain accurate global text information,a text classification model based on A-BiLSTM dual channel and optimized CNN was proposed.A-BiLSTM double-channel layer,which not only used the attention mechanism to focus on the part that contributes to the text classification,but used BiLSTM to extract the con-textual semantic information in the text.A-BiLSTM two-channel layer fused the output feature information of both to obtain high-level semantics.After A-BiLSTM double-channel layer,key local features were extracted with the strong learning ability of optimized CNN,and the final text feature representation was obtained.The classifier was used to output the category of text information.Experimental results show that the classification effects of this model are better than that of other comparison mo-dels,and it has good generalization ability.
关键词
文本分类/深度学习/双通道网络/注意力机制/双向长短时记忆网络/卷积神经网络/词向量模型Key words
text classification/deep learning/dual channel network/attention mechanism/BiLSTM/CNN/word vector model引用本文复制引用
基金项目
国家自然科学基金面上项目(61772449)
河北省自然科学基金青年项目(D2021402043)
邯郸市科技局项目(21422093285)
出版年
2024
NETL
NSTL
维普
万方数据