Predicting the popularity of micro-reviews: A Foursquare case study

We tackle the problem of predicting the future popularity level of micro-reviews, focusing on Foursquare tips, whose high degree of informality and briefness offer extra difficulties to the design of effective popularity prediction methods. Such predictions can greatly benefit the future design of content filtering and recommendation methods. Towards our goal, we first propose a rich set of features related to the user who posted the tip, the venue where it was posted, and the tip's content to capture factors that may impact popularity of a tip. We evaluate different regression and classification based models using this rich set of proposed features as predictors in various scenarios. As fas as we know, this is the first work to investigate the predictability of micro-review popularity (or helpfulness) exploiting spatial, temporal, topical and, social aspects that are rarely exploited conjointly in this domain.