Identification of Influential Visual Markers in Extremist YouTube Videos from the Viewpoint of Younger Viewers through Partial Least Square (PLS) Analysis

摘要

Videos related to extremists containing violent and graphic contents referred as 'extremist YouTube videos'. Abundance of these videos found on the YouTube despite prohibition by the community guideline. Extremist YouTube videos may vary in language, size format and purpose. However, the similarities are apparent on impressive messages that emotionally moved the viewers via embedded part that calls for visual attention - also known as visual markers. To date, little investigation has been done to identify visual markers that have influence on viewers' emotion. Eighty university students were recruited to rate their emotional responses upon watching 20 extremist YouTube videos during a Kansei evaluation. This paper will report on Partial least square (PLS) analysis of data obtained against 30 visual marker attributes converted into dummy variables. From PLS coefficient score, mean of range was calculated in order to identify the most influential visual markers according to particular emotion concepts. Six visual markers - profile, quantity, violent act, weaponry, logo and scene setting are identified as the most influential from viewpoint of younger viewers in evoking their emotion upon watching the videos. Nonetheless, the order of influence is slightly different for each of the emotion concept. The findings shall benefit academia and other stakeholders in understanding kind of emotion concepts evoked by which visual markers and future work on affective video classification.