Content recognition method and device based on double-attention neural network

The invention provides a content identification method and device based on a double-attention neural network, and the method comprises the steps: obtaining a communication operator corpus, generating a corresponding class label description for each class label through employing an automatic class label description generation mode, calculating a corresponding relation between text words and category labels in a corpus of a communication operator to obtain a label representation matrix; the method comprises the following steps: preprocessing work order information of a communication operator by using a text encoder to obtain preprocessed text representation; inputting the preprocessed text representation into a self-attention mechanism network to obtain a self-attention text representation; inputting the preprocessed text representation into a label attention neural network to obtain a label attention text representation; processing the self-attention text representation and the label attention text representat