Citance-Contextualized Summarization of Scientific Papers
This addresses the need for more informative and citation-aware summaries in scientific literature, though it is incremental as it builds on existing summarization techniques.
The authors tackled the problem of generating contextualized summaries of scientific papers that show the relationship between a paper and its cited references, resulting in a new approach that extracts citances and generates tailored summaries, evaluated on a dataset of 540K papers and 4.6M citances.
Current approaches to automatic summarization of scientific papers generate informative summaries in the form of abstracts. However, abstracts are not intended to show the relationship between a paper and the references cited in it. We propose a new contextualized summarization approach that can generate an informative summary conditioned on a given sentence containing the citation of a reference (a so-called "citance"). This summary outlines the content of the cited paper relevant to the citation location. Thus, our approach extracts and models the citances of a paper, retrieves relevant passages from cited papers, and generates abstractive summaries tailored to each citance. We evaluate our approach using $\textbf{Webis-Context-SciSumm-2023}$, a new dataset containing 540K~computer science papers and 4.6M~citances therein.