Chinese News Headline Classification Method Based on ERNIE

To solve the problems of semantic ambiguity and feature sparsity in news headline classification tasks, a news headline classification method based on ERNIE is proposed. Firstly, the input text is embedded by the pre-trained language model Enhanced Representation through Knowledge Integration (ERNIE) to obtain dynamic word vectors, solving the problem of polysemy. Then, we use TextCNN and set multiple sizes of convolutional kernels to extract different levels of semantic information from the text. Finally, softmax is used to normalize the output, and the category corresponding to the maximum probability value is taken as the classification result output. Experimental results show that this method can effectively solve the problems of sparse features and semantic ambiguity, and improve the classification performance of news texts.