Research on Chinese multi-documents automatic summarizations method based on improved TextRank algorithm and seq2seq

In this paper a two-stage automatic summarization model is proposed, which combines traditional method with deep learning method. In the first stage, this paper uses improved TextRank algorithm which combines with sentence weight to extract key sentences from multiple documents. In the second stage, a summary sentence is generated from the key sentences sequence based on the Seq2seq model. The experiments on LCSTS and self-constructed corpus show that the scores of the model in this paper of Rouge are all improved with character level input, which shows the effectiveness of the proposed method of this paper.