LAUSR

Today, online social networks (OSNs) constitute a major part of our lives and have, to a large extent, replaced traditional media for direct communication, as well as information dissemination and gathering. In the vast amount of posts that get published in OSNs each day, some posts do not draw any attention while others catch on, become viral, and develop as so-called buzzes. Buzzes are defined through their characteristics of immediacy, unexpectedness, and intensity. The early detection of buzzes is of vital importance for companies, public figures, institutions, or political parties—e.g., for the pricing of profitable advertising placement or the development of an appropriate social media strategy. While previous researchers developed systems for detecting trending topics, mainly characterized by their intensity, this is the first study to implement a buzz detection system (BDS). Based on almost 120,000 manually classified Facebook posts, we estimated and trained models for the BDS by applying various classification techniques. Our results highlight that, among other predictors, the number of previously passive users who then engage in the buzz post, as well as the number of likes given to the comments, are important. Evaluating the BDS over a five-month evaluation period, we found that these two classifiers perform best and detected over 97% of the buzzes.