The potential of emotions as predictors of news popularity on twitter

Abstract

News media has the main purpose to deliver information to the public, yet they also have to stay profitable. When news media uses social media as one of their platforms, popular contents would be beneficial for the media. Research has shown that contents with strong high-arousal emotions are more likely to capture a larger amount of interest on the Internet and are more likely to be disseminated. Despite the importance of emotions in the popularity of a content on the Internet, the research regarding the tweet popularity has not included emotions in the prediction models. This paper investigates whether or not emotions (compared to the binary sentiment of positive and negative) have the potential to predict tweet popularity regarding Coronavirus Disease 2019 (COVID- 19) news in various news media worldwide. First, the tweets are preprocessed by removing the symbols. Then, the emotions in each tweet is quantified based on a sentiment lexicon called NRC Affect Intensity Lexicon. Finally, the correlation between emotions and popularity (Retweet) is computed using Pearson Correlation Coefficient (PCC). Based on the tweets from 20 Twitter accounts of various news media worldwide, several findings are obtained. First, the numbers of Like and Retweet are highly linearly correlated. Second, several emotions (and their interactions) are linearly correlated (PCC > 0.1 or PCC < -0.1) in several news media. The correlations, however, are found to be different among news media accounts. Therefore, emotions are considered potential to become predictors of tweet popularity, as opposed to the mere positive or negative sentiment, that is commonly used in the existing research.