Identifying hidden geospatial resources in catalogues

Abstract

Geonetwork systems are one of the most popular geographic resources catalogues on the Web. One of the key elements of the successful of Geonetwork is the Geospatial metadata. Geonetwork use the direct and indirect spatial references from the metadata to process the spatial queries. However the inconsistencies between direct and indirect spatial references generate imprecise results, the invisibility of potential resources, and consequently the omission of many important data. The visibility of resources in a collection often depends on the consistency of the descriptions that help to find resources. Spatially inconsistent metadata records can hide resources and make them irretrievable. This paper presents an automatic method based on the combination of spatial clustering, reverse geocoding, and information retrieval techniques able to measure and detect geosemantic inconsistencies in metadata collections. Experimental results with a large metadata dataset about Geospatial Web Services show that the use of this method provides not only significant advantage in terms of accuracy, but also a gain of geosemantic insight into the metadata. Approach like this could be used by semantic technologies to draw connection from unstructured geospatial information to data from other structured geospatial information.