Evaluating the applicability of user location inference methodologies to increase the usability of Twitter data in event detection research scenarios

Abstract

Twitter data is applied in a diverse set of GIS research scenarios. The problem is that only a small portion of tweets in data sets used are geotagged (added a geographical coordinate to). Therefore geolocation inference methodologies (GIMs) have been designed over the years by researchers to increase the amount of geographical references in Twitter data sets through indirect means. Examples of indirect means are tweet content and social network of the user. Within the current academic framework there is no clarity what the applicability of different GIMs is in different GIS research scenarios. By conducting this thesis it was found that the increase of usability is a matter of compromise rather than an overall increase of data usability.