Automatic Text Summarization Using Deep Reinforcement Learning and Beyond

In the era of big data, information overload problems are becoming increasingly prominent. It is challengingfor machines to understand, compress and filter massive text through use artificial intelligencetechnology. The emergence automatic summarization mainly aims at solving problem ofinformation overload, it can be divided into two types: extractive abstractive. former finds somekey sentences or phrases from original combines them a summarization; latter needs acomputer understand content then uses readable language for human tosummarize key text. This paper presents two-stage optimization method forautomatic that abstractive summarization. First,a sequence-to-sequence model with attention mechanism trained as baseline generate initialsummarization. Second, updated optimized directly on ROUGE metric by using deep reinforcementlearning (DRL). Experimental results show compared model, Rouge-1, Rouge-2,and Rouge-L have been increased LCSTS dataset CNN/DailyMail dataset.