1. HTPosum:Heterogeneous Tree Structure augmented with Triplet Positions for extractive Summarization of scientific papers.
- Author
-
Zhu, Zhenfang, Gong, Shuai, Qi, Jiangtao, and Tong, Chunling
- Subjects
- *
TRIPLETS , *MULTICASTING (Computer networks) , *GRAPH connectivity , *TREES , *SCIENTIFIC models - Abstract
Currently, many summarization models are based on Transformer architectures, which regard text as a connected graph of words and tokens, thus fail to fully leverage the text's hierarchical structure. However, hierarchical structural information is crucial in understanding long documents, especially scientific papers. To exploit the structural information, we propose a H eterogeneous T ree structure augmented with T riplet P ositions for extractive Sum marization of scientific papers (HTPosum). The proposed model constructs a heterogeneous tree for each document, introducing section nodes and a global node to capture the documents' structural information. The tree is then augmented with a novel triplet position, which consists of three values that represent the depth, parent width position, and sibling width position of each node in a tree. The triplet position provides a unified view of the document structure. Experimental results show that the proposed model achieves state-of-the-art extractive performance on the PubMed and arXiv datasets compared with the latest scientific paper summarization models. • A heterogeneous tree is built to model scientific papers' hierarchical structure. • A triplet position and unique triplet position encoding method is proposed. • The proposed model achieves competitive results compared to latest baselines. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF