Uphill from here: Sentiment patterns in videos from left- and right-wing

Abstract

News consumption exhibits an increasing shift towards online sources, which bring platforms such as YouTube more into focus. Thus, the distribution of politically loaded news is easier, receives more attention, but also raises the concern of forming isolated ideological communities. Understanding how such news is communicated and received is becoming increasingly important. To expand our understanding in this domain, we apply a linguistic temporal trajectory analysis to analyze sentiment patterns in English-language videos from news channels on YouTube. We examine transcripts from videos distributed through eight channels with pro-left and pro-right political leanings. Using unsupervised clustering, we identify seven different sentiment patterns in the transcripts. We found that the use of two sentiment patterns differed significantly depending on political leaning. Furthermore, we used predictive models to examine how different sentiment patterns relate to video popularity and if they differ depending on the channel's political leaning. No clear relations between sentiment patterns and popularity were found. However, results indicate, that videos from pro-right news channels are more popular and that a negative sentiment further increases that popularity, when sentiments are averaged for each video.

The temporal evolution of a far-right forum

“…Sentiment extraction: Finally, we measured each post's sentiment using a sentiment look-up table in an algorithm that considered valence shifters (e.g., "hardly", "not") of the sentiment values. Valence shifters can change the polarity of a sentiment ("don't like" vs "like") or amplify ("really bad" vs "bad") and de-amplify ("barely exciting" vs "exciting") a sentiment [25,26]. Specifically, we built a context window of two words before and after a sentiment match and corrected the sentiment if valence shifters were present in the resulting 5-word context window.…”

“…Trajectory extraction: The individual temporal trajectories for the posts made on the forum per user were obtained by standardising the temporal progression to a scale from 1 to 100, and by applying a discrete cosine transformation to the post frequency scores [25,26,33]. 3 The forum engagement was scaled to a score ranging from 0 (minimal forum engagement) to 1 (maximal forum engagement).…”

Online influence, offline violence: language use on YouTube surrounding the ‘Unite the Right’ rally

“…The method for retrieving YouTube video transcripts follows the procedure of related research [49,50]. To retrieve the transcripts, a Python script was written using www.downs ub.com to obtain XML-encoded transcripts.…”

“…Videos that contained fewer than 100 words were not considered for analysis, following previous work on YouTube transcripts [49,50]. Using R software, each video was checked for English language and was excluded if it contained fewer than 50% English words.…”

“…Furthermore, when we selected videos for analysis only transcripts with more than 100 words and a pre-specified percentage of English words were retained. These decisions were guided by previous research [49,50] and our aim to retain only high-quality transcripts suitable for topic modelling. A full dataset without these filters applied is available for other researchers to experiment with other constraints.…”