A Text Classification Method Based on Graph Attention Networks
Yong Liu, Xiangnan Gou · 2021
With the rapid generation and dissemination of information data in modern society, intelligent processing of text classification is becoming more and more important. The Sequential and Graph-based deep learning models are often used in Natural Language Processing (NLP). The Sequential model usually uses Recurrent Neural Network (RNN), Convolutional Neural Network (CNN) and Bidirectional Encoder Representations from Transformers (BERT) The model performs natural language processing. The graph-based depth model uses the Co-occurrence relationship between texts to learn the characteristics of texts and texts for classification. In this paper, we use RNN to preliminarily calculate the features in the text as the node of the graph, construct a graph with the help of the modification relationship between texts, and then use the graph model to obtain the final text features used to predict the text category. The experiment was compared with a variety of methods through a variety of data sets, and the results showed that the method in this paper achieved better results on the text data set used for emotion classification, and the accuracy rate reached 82.03%.