Supplementary material online include: data link, python code and an instruction file needed to reproduce the results; an appendix containing additional structures and experiments we have tried. The web link is
Previous abstractive methods apply sequence-to-sequence structures to generate summary without a module to assist the system to detect vital mentions and relationships within a document. To address this problem, we utilize semantic graph to boost the generation performance. Firstly, we extract important entities from each document and then establish a graph inspired by the idea of distant supervision (