A Combined Extractive With Abstractive Model for Summarization

Aiming at the difficulties in document-level summarization, this paper presents a two-stage, extractive and then abstractive summarization model. In first stage, we extract important sentences by combining similarity matrix (only used for time) or pseudo-title, which takes full account of features (such as sentence position, paragraph more.). To coarse-grained from document, considers differentiation most document. The second stage is abstractive, use beam search algorithm to restructure rewrite these syntactic blocks extracted sentences. Newly generated summary serves pseudo-summary next round. Globally optimal pseudo-title acts final summarization. Extensive experiments have been performed on corresponding data set, results show our model can obtain better results.