Rauch, Erik and Bukatin, Michael and Baker, Kenneth (2003). ''A confidence-based framework for disambiguating geographic terms'', Proceedings of the HLT-NAACL 2003 workshop on Analysis of geographic references, Association for Computational Linguistics, pages 50--54
This article provides a set of excellent ideas regarding the disambiguation of geographic terms, using "positive" (agreeing) and "negative" (disagreeing) context. One of the main problems is that not all information required for disambiguating terms is locally available (in fact spatial references are even strongly non-local).
The authors present an approach considering closeness, population and concepts like nearby/enclosed by. Facilitating the local document context yields better results for geo/non-geo disambiguation (terms like Mr/Dr versus city of, mayor of, community, ...). In the end every place (p) is assigned a probability (p).