Extracting Location Information from Crowd-sourced Social Network Data

With millions of users worldwide, crowd-sourced social media data provide a valuable data source for events happening around the world. More specifically, microblogs, which are social networks that enforce short text messages, have a high popularity due to their availability as a mobile application and the practicality of short messages. Estimating the location of the events detected by following posts in microblogs have been the motivation of numerous recent studies. Extracting the location information and estimating the event location is a challenging task to maintain satisfactory situation awareness, especially for emergency cases such as fire or traffic accidents. Today, Twitter is among the most popular microblogging platforms, and there are recent research efforts aimed at detection of novel events online by following the Tweets. In order to analyze events, researchers generally focus on spatio-temporal features of the posts. Temporal features denote the time and ordering of posts, whereas spatial features are useful for location extraction or estimation. In this work, we present an overview on the process for toponym recognition and location estimation from microblogs.