Nowadays, on-line news agents post news articles on social media platforms with the aim to spread information as well as to attract more users and understand their reactions and opinions. Predicting the emotional influence of news on users is very important not only for news agents but also for users, who can filter out news articles based on the reactions they trigger. In this paper, we focus on the problem of emotional influence prediction of a news post on users before publication. For the prediction, we explore a range of textual and semantic features derived from the content of the posts. Our results show that terms is the most important feature and that features extracted from news posts' content allow to effectively predict the amount of emotional reactions triggered by a news post.