Towards tweet content suggestions for museum media managers

Cultural Heritage institutions are embracing social technologies in the attempt to provide an effective communication towards citizens. Although it seems easy to reach millions of people with a simple message posted on social media platforms, media managers know that practice is different from theory. Millions of posts are competing every day to get visibility in terms of likes and retweets. The way text, images, hashtags and links are combined together is critical for the visibility of a post. In this paper, we propose to exploit machine learning techniques in order to predict whether a tweet will likely be appreciated by Twitter users or not. Through an experimental assessment, we show that it is possible to provide insights about the tweet features that will likely influence its reception/recommendation among readers. The preliminary tests, performed on a real-world dataset of 19,527 museum tweets, show promising accuracy results.