Section-Aware Scientific Paper Summarization System Using Transformers

Abstract: The need for automatic summarizing systems has grown as the quantity of scientific papers has expanded. This paper provides a Section-Aware Scientific Paper Summarization System that generates structured summaries by combining transformer models like BART, T5, and LED with BM25-based extractive techniques. In order to improve readability and information access, the system divides summaries into sections titled Introduction, Methodology, Results, and Conclusion. ROUGE measures are used to assess performance, and the findings demonstrate that the suggested method successfully maintains important information while generating succinct and logical summaries.

Keywords: Scientific Paper Summarization, BART, T5, LED, BM25, ROUGE, Natural Language Processing.