Eliciting fuzzy location data from social media posts with Natural Language Processing

Social Media Platforms such as Twitter are collecting large volumes amounts of user generated content every day, much of which is location aware. Whilst there are clear use cases for data harvested from these platforms in research, harvesting and analysis of these datasets represents a substantial challenge. Existing work exploits geotagged or geocoded metadata collected alongside user generated content (and the challenges that accompany its use). However, the majority of user generated content lacks explicit locational information, but may still contain location information albeit in a less explicit, or “fuzzy”, form such as textual descriptions in the main body of the social media posting. This work explores the analysis of such datasets, and presents a novel generalisable methodolo