Geo-Coding for the Mapping of Documents and Social Media Messages
Abstract
Many places on the earth have the same name, so it is difficult to determine which written place is meant. This research aims to improve the precision of geo-coding by using natural language processing and machine learning techniques (SVM specifically). We used data that was already geo-coded the ACE Spatial ML data set, and a large tweet set in which tweets were selected for having GPS locations (that we could use to improve validity). The report details our methods for text and microtext.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 22, 2013
- Accession Number
- ADA595532
Entities
People
- Judith Gelernter
Organizations
- Carnegie Mellon University