Feature Extraction of News Text based on MMR Algorithm

Aiming at the problem that the current text classification neural network cannot fully extract the semantic structure feature information between words and sentences and sentences, a text feature extraction based on MMR algorithm is proposed. First, the MMR is used to extract the words and words, the sentence and sentence features information of the text respectively; secondly, the layered attention mechanism network layer is used to select the important words and sentences in the text respectively; finally, the MMR algorithm will be one by one. The text features obtained by layer extraction are used for text classification. The results show that the MMR algorithm can effectively extract the features of news text.