Characteristic Analysis of SNS Videos for Detection of Flaming Using Sentiment Analysis and Principal Component Analysis

With the spread of social networking services (SNS), the phenomenon known as flaming, in which slanderous and inflammatory comments are made on SNS, has increasingly become a social problem, especially on popular video sharing services such as YouTube. However, few studies have analyzed flaming by focusing on users’ emotions as they are expressed through video comments. We therefore propose a novel method for analyzing flaming comments on YouTube videos using a combination of sentiment analysis and principal component analysis. Using the proposed method, we analyzed comment data from both controversial and non-controversial videos on YouTube and show characteristics of flaming.