A text sentiment classification method based on graph convolution network
In order to better analyze the role of words and long-distance dependencies,and explain relevant syntactic constraints,this paper proposes a graph convolutional network model MH-GCN based on multi head attention mechanism and graph convolutional network.A graph conv-olutional network is established on the dependency tree of a sentence to utilize syntactic infor-mation and single word dependencies.Utilize multi head attention mechanism to learn relevant information of multiple representation subspaces,and use graph convolutional networks to ob-tain syntactic information and long-distance dependencies.Experiments have shown that the MHGCN model can effectively complete sentiment classification tasks and provide reference ba-sis for human-computer interaction,healthcare,and social media public opinion monitoring.
natural language processingsentiment classificationgraph convolution networkmultiple attention mechanismbilstm
NETL
NSTL
万方数据
