Analysis and Research of Abstractive Automatic Summarization Based on Sequential Facts

Automatic summarization is a task of converting text, and the summary result obtained should be able to accurately describe the facts that occurred in the original text. But so far, there are a lot of factual errors in the results obtained by generative summary models, resulting in low quality and poor readability. We believe that adding factual information in the encoding stage can effectively improve the readability of the summary and generate more accurate facts. To this end, we propose an abstractive summary model based on sequential facts and conduct experiments on the CNN/Daily Mail dataset. Experiments have proved that the integration of factual information can effectively improve the ROUGE value and factual accuracy of the summary.