Textgrounder
Start date: September 2010
Projected publication date: ongoing
TextGrounder is a system that processes texts to identify the places and times that are mentioned in them and disambiguates them to points on Earth or on the timeline. The system uses a number of natural language processing components, plus unsupervised machine learning methods for disambiguation. The output of the system can be used to create various geotemporal visualizations of texts and text corpora, for example, on Google Earth.
Textgrounder is the brainchild of Jason Baldridge (Linguistics, UT Austin), and is supported by the New York Community Trust. Matt Cohen and Jason Baldridge are co-PIs of a grant to develop Textgrounder for pedagogical uses from Liberal Arts Instructional Technology Services at UT Austin. We're trying to make it easy to pull geolocations from literary texts, focusing on frequently taught texts from nineteenth-century American literature.
See http://code.google.com/p/textgrounder/ for more information and a glimpse of the state of the project.