Revising geographical knowledge: A model for local belief change

Abstract

The revision problem is known to be a very hard problem. There is no revision algorithm which is able to handle a large amount of data. Geographical information systems (GIS) are characterized by a huge amount of data, often gathered from different sources of information, which are imperfect and whose quality can differ. Therefore GIS require the definition, and the use of belief revision. Literature on the topics shows that all known revision operators (i.e. from a global point of view), are unable to solve problems of this size. In this chapter, we show how to take advantage of the geographical context to define local revision operators that can be combined to handle the global revision problem. For this purpose, we define a postulate that may be assumed with geographical data - the containment assumption - and we show how this postulate can be captured by a new knowledge representation model, the G-structure model. Then we define a revision operation on this model, which can be run locally, and we apply this operation on a real experiment, with real data, which we succeed to process, correctly, though the global revision always failed. © 2010 Springer-Verlag Berlin Heidelberg.