Cutting Through the Comment Chaos: A Supervised Machine Learning Approach to Identifying Relevant YouTube Comments

Social scientists often study comments on YouTube to learn about people's attitudes towards and experiences of online videos. However, not all YouTube comments are relevant in the sense that they reflect individuals' thoughts about, or experiences of the content of a video or its artist/maker. Therefore, the present paper employs Supervised Machine Learning to automatically assess comments written in response to music videos in terms of their relevance. For those comments that are relevant, we also assess why they are relevant. Our results indicate that most YouTube comments are relevant (approx. 78%). Among those, most are relevant because they include a positive evaluation of the video, describe a viewer's personal experience related to the video, or express a sense of community among the video viewers. We conclude that Supervised Machine Learning is a suitable method to find those YouTube comments that are relevant to scholars studying viewers' reactions to online videos, and we present suggestions for scholars wanting to apply the same technique in their own projects.