Tracking geographical locations using a geo-aware topic model for analyzing social media data

Marianela García Lozano,Jonah Schreiber,Joel Brynielsson

doi:10.1016/j.dss.2017.05.006

Abstract

Tracking how discussion topics evolve in social media and where these topics are discussed geographically over time has the potential to provide useful information for many different purposes. In crisis management, knowing a specific topic's current geographical location could provide vital information to where, or even which, resources should be allocated. This paper describes an attempt to track online discussions geographically over time. A distributed geo-aware streaming latent Dirichlet allocation model was developed for the purpose of recognizing topics' locations in unstructured text. To evaluate the model it has been implemented and used for automatic discovery and geographical tracking of election topics during parts of the 2016 American presidential primary elections. It was shown that the locations correlated with the actual election locations, and that the model provides a better geolocation classification compared to using a keyword-based approach.

Highlights

The rapid growth of social media has enabled millions of people to express themselves and reach wide audiences
Topic models and latent Dirichlet allocation (LDA) Topic models assume that the content of documents is governed by abstract topics which are hidden
A prototypical tool is provided with which the topics and trending locations of streaming media can be automatically discovered

Summary

Introduction

The rapid growth of social media has enabled millions of people to express themselves and reach wide audiences. The geo-positioning methods have employed a number of indirect indicators such as looking at the social network, the style of writing (choice of word, etc.), or explicit location references given in the text content [3,4,5,6]. Combining these types of location indicators with metadata such as timezone, language, and even posting time, has resulted in fairly good localization of both user residence and tweet post location [7,8]. The results of these efforts usually do not differentiate between the account holder’s current or past locations, and they do not focus on the topical correlation

Objectives

Methods

Results

Discussion

Conclusion