Text classification method based on graph convolutional neural network with self-attention mechanism

Abstract

The invention provides a text classification method based on a graph convolutional neural network with a self-attention mechanism, and relates to the technical field of artificial intelligence and information technologies. The method comprises the steps of: obtaining and storing multiple to-be-classified texts in a corpus, meanwhile, performing word segmentation processing on the to-be-classified texts to obtain a text word segmentation sequence, and performing preprocessing; then obtaining a self-attention mechanism matrix of the text word segmentation sequence by using a self-attention mechanism; constructing a graph network structure for all texts; preprocessing the graph network structure, and calculating and mormalizing a Laplacian matrix of the graph; constructing and training a convolutional neural network model of the graph on the basis of the Laplacian matrix of the graph; and finally, obtaining a text classification result through a Softmax classifier. According to the classification method, the semantic information correlation between the texts can be better captured, so that the implicit relationship in the text information is better expressed, and accurate classification of the texts is realized.

Description

technical field [0001] The invention relates to the technical fields of artificial intelligence and information technology, in particular to a text classification method based on a graph convolutional neural network with a self-attention mechanism. Background technique [0002] With the rapid development of deep learning technology, the data scale shows an explosive growth trend. More and more researchers apply deep learning and neural network methods to the field of graph network structure, which promotes the rapid development of deep learning research field. Graph neural network is a kind of method based on deep learning to process graph network structure, and it has better performance and interpretability. In just a few years, in view of the wide application of neural networks in the fields of images and texts, some researchers have tried to combine neural network methods with graph network structures, and graph neural network research has gradually become an upsurge in t...

AI Technical Summary

Problems solved by technology

Although these methods have their own advantages, it is difficult to guarantee the optimal model for the overall classification effect

Images