Paper update: Geotagging One Hundred Million Twitter Accounts with Total Variation Minimization

Press coverage (updated 2015-05-02):

Earlier this week I updated a paper on the arxiv: http://arxiv.org/abs/1404.7152. The paper appeared a few months ago at IEEE BigData 2014 and was the subject of all my talks last year. This update syncs the arxiv paper with what was accepted for the conference. Here’s the abstract:

Geographically annotated social media is extremely valuable for modern information retrieval. However, when researchers can only access publicly-visible data, one quickly finds that social media users rarely publish location information. In this work, we provide a method which can geolocate the overwhelming majority of active Twitter users, independent of their location sharing preferences, using only publicly-visible Twitter data.

Our method infers an unknown user’s location by examining their friend’s locations. We frame the geotagging problem as an optimization over a social network with a total variation-based objective and provide a scalable and distributed algorithm for its solution. Furthermore, we show how a robust estimate of the geographic dispersion of each user’s ego network can be used as a per-user accuracy measure which is effective at removing outlying errors.

Leave-many-out evaluation shows that our method is able to infer location for 101,846,236 Twitter users at a median error of 6.38 km, allowing us to geotag over 80% of public tweets.