ScisummNet: A Large Annotated Corpus and Content-Impact Models for Scientific Paper Summarization with Citation Networks

TLDR

Scientific article summarization is difficult because large annotated corpora are lacking and summaries should reflect the paper’s impact on the research community. This work introduces a large, manually annotated corpus for computational linguistics papers and proposes hybrid summarization methods that combine author highlights with citation-based impact signals. The authors built the corpus, designed hybrid summarization models, and ran experiments showing that these models outperform plain abstracts and traditional citation summaries. The corpus and hybrid methods establish a new framework for scientific paper summarization research.

Abstract

Scientific article summarization is challenging: large, annotated corpora are not available, and the summary should ideally include the article’s impacts on research community. This paper provides novel solutions to these two challenges. We 1) develop and release the first large-scale manually-annotated corpus for scientific papers (on computational linguistics) by enabling faster annotation, and 2) propose summarization methods that integrate the authors’ original highlights (abstract) and the article’s actual impacts on the community (citations), to create comprehensive, hybrid summaries. We conduct experiments to demonstrate the efficacy of our corpus in training data-driven models for scientific paper summarization and the advantage of our hybrid summaries over abstracts and traditional citation-based summaries. Our large annotated corpus and hybrid methods provide a new framework for scientific paper summarization research.