GoSum: Extractive Summarization of Long Documents by Reinforcement Learning and Graph Organized discourse state

11/18/2022
by   Junyi Bian, et al.
1

Handling long texts with structural information and excluding redundancy between summary sentences are essential in extractive document summarization. In this work, we propose GoSum, a novel reinforcement-learning-based extractive model for long-paper summarization. GoSum encodes states by building a heterogeneous graph from different discourse levels for each input document. We evaluate the model on two datasets of scientific articles summarization: PubMed and arXiv where it outperforms all extractive summarization models and most of the strong abstractive baselines.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset