Enhancing Identification of Scholarly Reference on YouTube

Abstract

Scholarly communication through YouTube videos has been increasing. Although Altmetric provides the dataset on such references, its coverage is unclear, and it does not contain the original external links in each video. Further investigation is needed to understand the characteristics of scholarly references as external links in YouTube videos. To address this gap, we propose a method to identify scholarly references by searching for domain names and building a dataset by applying this method. Subsequently, we compare this dataset with the Altmetric dataset and analyze the external link characteristics. Using the proposed method and targeting six types of domain names, we identified approximately 480,000 references among 230,000 videos posted on 55,000 channels. Notably, over half of these references were not covered by the Altmetric dataset, resulting in a 150% increase in the number of references when combining the dataset constructed by the proposed method with the Altmetric dataset, compared to the Altmetric dataset alone. Regarding external links, PubMed and DOI links were prominent; however, a substantial number of direct links to publisher platforms were observed. Most channels and videos contained external links to a single platform, scattered across each platform. The method proposed in this study is helpful for identifying and analyzing scholarly references on YouTube. In addition, the findings on external link characteristics raise concerns about the long-term accessibility and fact-checking of information sources for YouTube video content.