A Presentation Generation Method Combining Topic Segmentation and Text Summarization

Disseminating scholarly achievements through presentations has become a common practice; however, manually creating presentations is burdensome. Current research mainly employs text summarization to extract or generate sentences from scientific papers for automatic slide generation, but the unique structural features and language style of slides are often overlooked. In this paper, we propose a method that combines topic segmentation and text summarization to generate clearly organized slides that reflect the logical structure of the scientific paper while containing less redundancy. Our method first segments the main text into multiple topics based on section structures to construct a hierarchy of slides before extracting significant textual units using text summarization for slide generation based on these topics. Experiment conducted on public datasets suggests that our approach outperforms baseline methods in generating slides according to ROUGE metrics while also reflecting the logical flow of information presented in scientific papers.