Unverified information can spread on the web and influence public opinion before it is eventually verified as true or false. Such rumors eventually evolve during reiterated transmission, and often accompany fake news. The use of social media has made this phenomenon something new with respect to what sociology has traditionally studied.

“Can rumors be scientifically understood and controlled?” According to Zubiaga et al., the question is yet to be answered. The authors examine automatic detection and the analysis of rumors; they consider about 200 papers and focus on rumors that circulate for long periods, and on their emergence from breaking news. The topics covered are accessing social media to collect and annotate data, the detection and tracking of rumors, dataset generation, rumor stance classification, and veracity classification. Data mining and natural language processing methods are used.

The article presents a large overview of research results, well organized into 11 sections. Available datasets for every topic are also reported. The review is very useful for people working in the vast area of rumor detection; it is much too detailed for didactic purposes.