Research On Text Classification Method Based On Graph Convolutional Network

With the development of the Internet,massive amounts of data are generated every day with different scales and forms.Text classification technology can classify massive amounts of data efficiently and accurately,making it convenient for users to obtain data.As a key sub-task in natural language processing,text classification has achieved good results.However,most models only represent the text in a vectorized representation.It is considered that the texts are independent of each other.In most citation network datasets The above models did not achieve the best results.Therefore,the paper first proposes a method of text representation,which not only considers the correlation between words and words,words and text,but also expresses the connection between text and text in the graph based on sparse representation and collaborative representation.The connection between words and words,words and text,text and text is reflected on the edges between the nodes in the graph.After obtaining the method of text graph representation,a model of text classification based on a two-layer graph convolutional network is proposed,which transforms the traditional text classification problem into a graph node classification problem.At the same time,the influence of the number of graph convolution layers in the model on the classification effect is studied,Applying the model to multiple datasets,the effect of text classification on most datasets has been improved,but for larger datasets,the classification effect is not optimal.To this end,two attention mechanisms are added to the text classification model of the two-layer graph convolutional network,namely V-attention mechanism,which is used to obtain the key features of the text,and E-attention mechanism to extract the high-dimensional features of the edges in the graph.By adding these attention mechanism,the performance of the model on large-scale datasets has been improved.The text representation method proposed in this paper integrates the relationship between texts into the training process of the graph convolutional text classification model.Experiments verify the influence of the attention mechanism on the model.Finally,the experimental results on multiple data sets show that,The model proposed in this paper plays a positive role in improving the text classification effect.