Positioning localities from spatial assertions based on Voronoi neighboring

Abstract

With the rapid development of Internet, much spatial information contained in non-structured or semi-structured documents is available on the World Wide Web. In such documents, localities are always textually described using spatial relationships and named places, instead of numerical coordinates. Hence, extracting positional information from locality descriptions is an important task. In this paper, we bridge two aspects of locality descriptions, namely generating locality descriptions and positioning localities, and provide a method to compute probability density according to the selection probability of a reference object to describe the position of the target object. Refinement operation on uncertainty field is used to deal with locality description involving multiple reference objects. Three metrics are introduced to measure the results of positioning localities. We choose the mixed selection probability function based on Euclidean distance and Voronoi stolen-area to compute probability density function. Finally, we use three cases to demonstrate the proposed methods.