Homing in on Twitter users: Evaluating an Enhanced Geoparser for User Profile Locations

摘要: Twitter-related studies often need to geo-locate Tweets or Twitter users, identifying their real-world geographic locations. As tweet-level geotagging remains rare, most prior work exploited tweet content, timezone and network information inform geolocation, else relied on off-the-shelf tools geolocate users from location in user profiles. However, such metadata is not consistently structured, causing fail regularly, especially if a string contains multiple locations, locations are very fine-grained. We argue that profile (UPL) be treated as distinct types of which differing inferences can drawn. Here, we apply geoparsing UPLs, demonstrate how task performance improved by adapting our Edinburgh Geoparser, was originally developed for processing English text. present detailed evaluation method results, including inter-coder agreement. the optimised geoparser effectively extract geo-reference at different levels granularity with an F1-score around 0.90. also illustrate geoparsed UPLs international trade country-level sentiment analysis.