A comprehensive analysis of tweet content and its impact on popularity

By the appearance of the online social networks, ordinary people have gained more chance to make and publish content. However, for audiences, as the number of these shared contents grows, the importance of detecting important and related ones increases. So, a significant question is how much would a shared content become popular among audiences, regardless of the source of content? In this paper, we investigate this question by performing a comprehensive analysis of tweet content and studying its impact on popularity of tweet. Here, the number of retweets of tweet is used as a popularity measure. We show that tweets with “social” content, have in general more chance of popularity due to their attraction for society. In contrast, tweets with “individual” content, have little chance to get popular. We collect a fair data set of tweets. In order to do more detailed investigation and access the semantic features, we set an annotation and labeling process. We analyze the informativeness of content-based features and use them to train predictive models. The results clearly show the importance of content-based features. They specifically support this idea that specifying whether a tweet is speaking about an individual or social subject, is the most informative content-based feature to predict the popularity i.e. the number of retweets.