Text Summarization for Research Papers using Transformers

Scientific research frequently begins with a thorough review of the body of previous work, which includes a wide range of publications. This study process might be shortened by automatically summarizing scientific publications, which would be of great use to researchers. Because scientific papers have a different structure and require citation phrases, summarizing them presents different issues than summarizing other forms of literature. Furthermore, algorithmic pseudocode, figures, tables, and other elements not typically present in generic text, often contain valuable content that is not readily visible in plain language. Our Research Paper assistant presents a comprehensive investigation into the development and analysis of the summary of a complex research paper leveraging a robust sequential model (seq2seq) using the Transformer model. We are performing a comparison between Extractive and Abstractive approaches as well to prove that this work highlights the necessity of Abstractive summary methodologies in research paper summarization, acknowledging the drawbacks of extractive methods. The complex structure and substance of research publications might be missed by extraction approaches, which results in a loss of context and coherence. Conversely, Abstractive summarization techniques provide an answer by producing unique, short, and cohesive text in addition to picking and rearranging sentences. In order to efficiently condense the huge amount of information and insights contained in research articles while preserving readability and informativeness, this shift towards Abstractive summarization is essential.