An Extractive Text Summarization Based on Reinforcement Learning

Abstract

Abstract: In recent years, with the rapid development of network information technology, network text information also presents an explosive growth trend. As an efficient information processing technology in the digital age, text summarization can bring the advantage of focusing on key information in all directions in massive text information. However, text summarization is still faced with some problems such as difficulty in extracting long text and information redundancy. Therefore, combining with the deep learning framework, this paper proposes an extractive text summarization that uses reinforcement learning to optimize the long text extraction process and uses the attention mechanism to achieve the effect of redundancy removal. On CNN/Daily Mail datasets, the automatic evaluation shows that our model outperforms the previous on ROUGE, and the ablation experiment proves the effectiveness of the de-redundant attention module.