Geolocating tweets via spatial inspection of information inferred from tweet meta-fields

Research output: Contribution to journalArticlepeer-review


In the last 10, years the Twitter social network has become a robust messaging platform. To date, Twitter has more than 500 million users worldwide. A given tweet can contain the geographic location of the transmitting device if the geolocation services are activated or the IP address can be geocoded, which is applicable to only 1–3% of all tweets. Nevertheless, tweets also contain more than 40 other meta-fields, some of which are inserted by the user and may include spatial information that can be tapped to infer locations associated with the tweet. This study implemented the publicly available GeoNames and Open Street Map (OSM) datasets in conjunction with curated dataset of 2001 tweets from Israel. With both Geonames and OSM, the inference of the tweets’ geographic locations was implemented using the meta-fields of the text (resulting in 600 geolocated tweets), the user location (857 tweets), and the user description (425 tweets). The inferred locations were then spatially examined to verify if they can serve as potential proxies. It was found that the distance between the inferred locations using the text and user-location meta-fields is well correlated with the distance between their midpoint and the device location. Thus, it may indicate the actual device location as well the location of the phenomenon described in the tweet.
Original languageEnglish
Article number102593
JournalInternational Journal of Applied Earth Observation and Geoinformation
StatePublished - 2021

Bibliographical note

Publisher Copyright:
© 2021 The Author


  • GeoNames
  • Geolocation
  • Open Street Map
  • Spatial inspection
  • Twitter

ASJC Scopus subject areas

  • Computers in Earth Sciences
  • Earth-Surface Processes
  • Management, Monitoring, Policy and Law
  • Global and Planetary Change


Dive into the research topics of 'Geolocating tweets via spatial inspection of information inferred from tweet meta-fields'. Together they form a unique fingerprint.

Cite this