HetTreeSum: A Heterogeneous Tree Structure-based Extractive Summarization Model for Scientific Papers

作者：

Highlights：

• A novel tree structure-based extractive summarization model is proposed for scientific papers.

• Structural information and inter-sentence relations are captured with graph transformer networks.

• A Bottom-Up-Top-Down iterative updating strategy is developed for mutual reinforcement.

• Our method achieves state-of-the-art performance in PubMed and arXiv datasets.

摘要

•A novel tree structure-based extractive summarization model is proposed for scientific papers.•Structural information and inter-sentence relations are captured with graph transformer networks.•A Bottom-Up-Top-Down iterative updating strategy is developed for mutual reinforcement.•Our method achieves state-of-the-art performance in PubMed and arXiv datasets.

论文关键词：Scientific paper summarization,Heterogeneous tree structure,Inter-sentence relations,Structural information,Iterative updating strategy

论文评审过程：Received 10 April 2022, Revised 15 July 2022, Accepted 30 July 2022, Available online 5 August 2022, Version of Record 17 August 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118335