摘要
为提高中文评教文本情感分析的准确率并挖掘文本中深层次语义情感关联,文章提出了一种基于ERNIE-BiAGRU-CapsNet的中文评教文本情感分析模型。其中,ERNIE通过词嵌入层获取文本中的隐含关系及深层语义动态词向量;BiAGRU在BiGRU更新门加入了注意力机制,以深入挖掘文本上下文语义表征信息;针对提取的上下文语义特征,CapsNet旨在进一步挖掘其局部语义特征。通过在实际中文评教文本数据中进行实验验证,模型分类输出的精确率高达96.77%,其分类性能较同类深度学习模型更优。
To improve the accuracy of sentiment analysis in Chinese teaching evaluation texts and explore deep semantic sentiment associations in texts,this article proposes a Chinese teaching evaluation text sentiment analysis model based on ERNIE-BiAGRU-CapsNet.Among them,ERNIE obtains implicit relationships and deep semantic dynamic word vectors in the text through the word embedding layer.BiAGRU has added attention mechanism to the BiGRU update gate to deeply explore the semantic representation information of text context.CapsNet aims to further explore the local semantic features of the extracted contextual semantic features.Through experimental verification on actual Chinese teaching evaluation text data,the accuracy of the model's classification output is as high as 96.77%,and its classification performance is better than similar deep learning models.
作者
仇全涛
张旭初
QIU Quantao;ZHANG Xuchu(School of Applied Technology,Anyang Preschool Teachers College,Anyang,Henan 455000,China)
