News text abstract generation method based on deep learning

A news text abstract generation method based on deep learning comprises the following steps: 1) crawling a news title and a news text to obtain original data; 2) preprocessing the data to obtain a data format required by the model; 3) constructing an extraction type abstract generation system and a generation type abstract generation system by using a Bert model as a core; 4) extracting sentences with high relevancy with a title from the input text by the extraction type model, wherein the sentences are used as guide signals of the generative model; 5) inputting the input text and the guidance signal into the generative model to generate a final abstract; and 6) comparing and evaluating the generated abstract with a reference abstract. According to the method, abstract generation is performed on the news text in combination with the extraction type abstract model and the generation type abstract model, so that the lengthy text can be simplified, and key information in the text can be quickly obtained. 一种基于深度学