A comparison of geometric approaches to assessing spatial similarity for GIR

This research compares the geographic information retrieval (GIR) performance of a set of logistic regression models with those of five non-probabilistic methods that compute a spatial similarity score for a query-document pair. All methods are applied to a test collection of queries and documents indexed spatially by two convex conservative geometric approximations: the minimum bounding box (MBB) and the convex hull. In the comparison, the tested logistic regression models outperform, in terms of standard information retrieval recall and precision measures, all of the non-probabilistic methods. The retrieval performance achieved by the logistic regression models on MBB approximations is similar to that achieved by the use of the non-probabilistic methods on convex hulls. Although these results are valid only for the test collection used in this study, they suggest that a logistic regression approach to GIR provides an alternative to the use of higher-quality geometric representations that are more difficult to obtain, implement, and process. Additionally, this research demonstrates the ability of a probabilistic approach to effectively incorporate information about geographic context in the spatial ranking process.