Evolving Approaches to Place Tagging in Social Media

Abstract

As social media platforms are maturing, their approach to location tagging increasingly emphasizes ontologies of named places: cities, neighborhoods, and predefined points of interest. As recently as 2015, millions of Twitter posts per month were tagged with exact coordinates, but these have all but vanished in the US user base. This paper evaluates the changing location content of public posts on Twitter (including cross-posts from Instagram and Foursquare Swarm), and the implications for research that relies on these data sources, using a 36-month sample of 18 million posts from the western US between 2015 and 2018. Place tags range in granularity from country-level to points of interest; the majority of tagged Twitter posts are associated with a city, while points of interest are the most common choice on Instagram. Varying granularity makes it hard to observe end-to-end local travel using these data, but the richer semantic content is an advantage for studying land use or discretionary activity patterns. The shift from exact coordinates to named places provides better location privacy and a better user experience and is aligned with a variety of business incentives. It is likely to be an enduring standard.