Investigating the opinions distribution in the controversy on social media

The research works that investigate controversies on social media seek to answer questions such as the following: Is the topic controversial? What opinions do the users hold? What is the size of each opinion group? Researchers can learn public points of view through the lens of the controversy. Unlike prior work, this study intends to investigate the distribution of opinions in the controversy via sentiments hidden behind rich user-generated content that is produced from interactions among users. To meet the challenge, this study develops an analytical framework that combines social network analysis and sentiment analysis. The framework constructs an emotional social network from replies and retweets on social media, and subsequently employs the simulated annealing algorithm to partition nodes in the network into three emotional factions, including proponent, opponent and neutral groups. Consequently, an opinion distribution in the controversy can be built. The framework also allows calculating a score based on the distribution to measure the controversy. We conducted experiments on a public dataset and a synthetic dataset. The experimental results demonstrate that the framework can perform the task well. We also employed the framework to explore Chinese movies and obtained several interesting findings, which include the following: (1) Both the very high-rated movies and the very low-rated movies have much lower controversy. (2) Almost all of the high box-office movies have high controversy. (3) A dominant positive image and high controversy are two salient characteristics of the high box-office movies.