Presentation styles in academic videos: Unveiling their relationships with scholarly impact and research novelty

Academic videos (AcVs) have emerged as a critical medium in multi-modal scholarly communication. Authors’ presentation styles in AcVs, operationalized as multimodal design strategies through temporal allocation, spatial distribution, and presentation speed across modality types and functional structures, provide valuable insights into how scientific knowledge is communicated and disseminated. However, existing studies have primarily focused on static, text-based communication, leaving multi-modal videos largely unexplored. To address this gap, we develop a dedicated analytical framework that applies AI techniques to understand and analyze AcVs. Furthermore, we introduce novel metrics to characterize presentation styles across temporal, visual, and speed dimensions. We also measure scholarly impact using citation-based indicators and assess research novelty using both embedding-based similarity and large language model (LLM) evaluations. Drawing on a dataset comprising 438 papers from ACL 2021 and their corresponding videos, we conduct a systematic analysis of authors’ presentation styles and examine their associations with both scholarly impact and perceived novelty. The results indicate that AcVs prioritize methodological and empirical sections through longer, multi-modally coordinated presentation. Our regression analyses further reveal that structured and balanced delivery is positively associated with higher impact, while visually rich and well-paced presentations, particularly in the method and experiment & conclusion sections, are linked to greater research novelty. This study represents the first attempt to quantitatively analyze presentation styles in AcVs and examine their associations with scholarly impact and novelty.