CQSumDP: A ChatGPT-Annotated Resource for Query-Focused Abstractive Summarization Based on Debatepedia

• DOI: 10.48550/arxiv.2305.06147
• Publication Date: 2023-01-01
• Authors: Md Tahmid Rahman ...; Mizanur Rahman; Israt Jahan; Enamul Hoque; Jimmy Xiangji Huang
• Keywords: Relevance; Benchmark (surveying); Multi-document summarization

Debatepedia is a publicly available dataset consisting of arguments and counter-arguments on controversial topics that has been widely used for the single-document query-focused abstractive summarization task in recent years. However, it recently found this limited by noise even most queries do not have any relevance to respective document. In paper, we present methodology cleaning leveraging generative power large language models make suitable summarization. More specifically, harness generation capabilities ChatGPT regenerate its queries. We evaluate effectiveness proposed annotated version using several benchmark demonstrate newly outperforms original terms both query as well summary quality. will cleaned available.