An engaging video can foster knowledge and understanding in learners in an enjoyable manner. A large number of videos on almost any topic can be freely sourced online. Watching all of them to select the most suitable ones for a course curriculum, however, can be impractical. Our study presents a semi-automated method to facilitate and accelerate a selection from social media videos. The method was examined using an initial list of 200 YouTube videos on climate change. Natural language processing of the Youtubers’ comments provided an effectiveness label for each video. Effective videos were defined as videos that elicited comments where the majority concurred with the effects of climate change. The effectiveness labels were manually rated by independent researchers who read the comments. The labels were then used to train a machine learning model. This model used the text of all comments and predicted the label for each video. The most prevalent words and phrases were extracted from comments to shed insight into the topic discussed in their texts. The method predicted 61 videos as effective, with an accuracy of 82%. The most frequent words and phrases in the comments explicitly addressed climate change issues. This preliminary proof of concept will be further pursued to provide more automation to the selection process, improve its accuracy and explore the feasibility of generalising the method to other topics. A publicly accessible online tool that implements the methodology is provided.

Authors: Vered Aharonson, Jared Joselowitz, S?awomir Nowaczyk

Published in: Canada International Conference on Education, 2024

• Date of Conference: 23-25 July, 2024
• DOI: 10.20533/CICE.2024.0052
• Electronic ISBN: 978-1-913572-65-5
• Conference Location: Toronto Metropolitan University, Toronto, Canada