Geolocation Prediction in Twitter Using Social Networks: A Critical Analysis and Review of Current Practice

By David Jurgens, Tyler Finethy, James McCorriston, Yi Tian Xu, Derek Ruths. 2015

231 citations recorded by the source corpus.

Geolocated social media data provides a powerful source of information about place and regional human behavior. Because little social media data is geolocation-annotated, inference techniques serve an essential role for increasing the volume of annotated data. One major class of inference approaches has relied on the social network of Twitter, where the locations of a user's friends serve as evidence for that user's location. While many such inference techniques have been recently proposed, we actually know little about their relative performance, with the amount of ground truth data varying between 5% and 100% of the network, the size of the social network varying by four orders of magnitude, and little standardization in evaluation metrics. We conduct a systematic comparative analysis of nine state-of-the-art network-based methods for performing geolocation inference at the global scale, controlling for the source of ground truth data, dataset size, and temporal recency in test data. Furthermore, we identify a comprehensive set of evaluation metrics that clarify performance differences. Our analysis identifies a large performance disparity between that reported in the literature and that seen in real-world conditions. To aid reproducibility and future comparison, all implementations have been released in an open source geoinference package.

Subfields: Epidemiology, Geography, Planning and Development, Transportation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Epidemiology

In Geography, Planning and Development

In Transportation

Citations in the corpus

Incorporating Dialectal Variability for Socially Equitable Language Identification (2017, methodology): “Individuals’ locations are inferred using the method of Compton et al. (2014) as implemented by Jurgens et al. (2015).”
Collective aspects of privacy in the Twitter social network (2018, methodology): “For example, Twitter data can be used to predict user locations [13, 14] as well as gender, age, and political orientation of users [15].”
You Are Who You Know and How You Behave: Attribute Inference Attacks via Users' Social Friends and Behaviors (2016, background): “Existing attribute inference attacks can be roughly classified into two categories, friend-based [18, 27, 46, 39, 15, 33, 11, 26, 20, 21, 23] and behavior-based [42, 29, 8, 25].”
Fake news on Twitter during the 2016 U.S. presidential election (2019, methodology): “We mainly relied on a gazetteer approach (33), in which we matched the text field against lists of known domestic and foreign cities, states, countries, and 4”
How Twitter is Changing the Nature of Financial News Discovery (2016, methodology): “This decision can be aided by methods that contextualize tweets [35], such as displaying inferred location [16, 20, 47, 9, 7, 28, 39, 14, 13] or topic context [29, 33].”
Semi-supervised User Geolocation via Graph Convolutional Networks (2018, background): “Network-based models fail to geolocate disconnected users: Jurgens et al. (2015) couldn’t geolocation 37% of users as a result of disconnectedness.”
Predicting Twitter User Demographics from Names Alone (2018, background): “Numerous existing systems automatically infer missing demographics, such as gender, ethnicity, age and location (Mislove et al., 2011; Burger et al., 2011; Culotta et al., 2015; Pennacchiotti and Popescu, 2011; Rao et al., 2010; Jurgens et al., 2015; Dredze et al., 2013; Rout et al., 2013).”
Everyone's Invited: A New Paradigm for Evaluation on Non-Transferable Datasets (2015, methodology): “As such, a full comparison between methods has been effectively prevented unless the computational models themselves are recreated and tested in one place in a setting that has access to comparable amounts of data as used in the original experiments, as was done in Jurgens et al. (2015).”
Geolocation for Twitter: Timing Matters (2016, methodology): “Geolocation work has mostly focused on Twitter, since tweets are readily accessible and true location available from user geocoded tweets (inter alia (Eisenstein et al., 2010; Han et al., 2014; Rout et al., 2013; Compton et al., 2014; Cha et al., 2015; Jurgens et al., 2015; Osborne et al., 2014; Dre”
The Geography and Importance of Localness in Geotagged Social Media (2016, methodology): ““Geometric Median” Localness Metric The geometric median metric [23] has been most commonly used in the geolocation inference literature to assign a home location to users.”