23-06-2023 дата публикации
Номер: CN116304061A
Принадлежит:
The invention discloses a text classification method and device based on hierarchical text graph structure learning and a medium, and the method comprises the following steps: S1, carrying out the preprocessing of a training set text according to three linguistic features, and obtaining three graph structure matrixes; s2, carrying out edge level graph structure learning to obtain three edge vectors; s3, redundancy is removed, and three text edge vectors are obtained; s4, performing weighted summation to obtain a text graph structure representation; s5, processing by adopting a graph convolutional neural network, and generating a graph-level text representation through a graph pooling layer; and S6, softmax classification is carried out, and the category with the maximum probability is a final classification result. The method has the advantages that the training set text is preprocessed by adopting three linguistic features, and a text classification problem is converted into a graph classification ...
Подробнее