Summary
These researchers chose to analyze the videos and comments of right-wing YouTube channels. To do this, the researchers looked at their language, topics, and biases. They analyzed over 7,000 videos and 17 million comments in their process. Using these, they observed similarities and differences between the videos and compared them to a normal dataset using layers. The researchers sought to answer 1) is the presence of hateful content more frequent in right wing channels and 2) how the comments compared to the video content in terms of hate and discrimination. The researchers discovered a general negativity in right wing videos, anti-muslim rhetoric in right wing communities, and negative words in both right wing videos and comments.
Reflection
I found the research article to be fairly interesting, because it covered an interesting political topic, but didn’t capture any results that were surprising. While I understand that it isn’t always fair to claim that research results are unsurprising, a major point to this paper is that it finds that YouTube comments are negative in right-wing communities. This should come as no surprise to anyone, since YouTube comments are already very negative and mixing political sentiment will only worsen the issue.
I found it odd that researchers decided that a future work item would be handling negations in text. This is something that has become fairly commonplace in natural language processing today. Any NLP library you come across should be able to propagate negations without issue. This might be an oversimplification of what they are talking about, but I don’t think that this change would be a difficult extension of their work.
Another easy extension of their work would be comparing their results to a dataset on a far left subcommunity. There are plenty to choose from, though likely not ones as ferocious as the far right in terms of language. Even still, observing the degree of difference would be very interesting. I bet that both groups cover similar topics and sentiment, but would have a stark difference in the type of language used.
