Personality detection method based on hypergraph attention mechanism

The invention discloses a personality detection method based on a hypergraph attention mechanism. The method comprises the following steps: collecting a post text of a user on a social media; then obtaining a text feature vector of the post text through a BERT pre-training language model so as to learn semantic features of the text; further obtaining a high-order semantic feature vector, and obtaining high-order semantic information of the post by adopting a hypergraph attention mechanism; and then, constructing a graph convolutional network, taking the text feature vector and the high-order semantic feature vector as nodes, constructing edges between the nodes through cosine similarity calculation, and learning through the graph convolutional network to obtain a final user personality representation. According to the method, higher-quality text representation is obtained through a hypergraph attention mechanism, the performance deficiency of an existing text representation method in processing the long-dista