An increasing number of thematic datasets are published as RDF graphs and linked to other datasets by identifying equivalent resources in other relevant datasets. Among the set of properties usually used as data linking criteria, geolocation (addresses, locations, coordinates ) remains one of the most commonly used.
However, resources that actually refer to complex topographic features are generally described by very simple geolocation properties, such as a position dened by coordinates (long, lat). On the other hand, geographic reference datasets provide more precise geometric information about geographic features. Interlinking thematic linked open datasets with geographic reference datasets would enable us to take advantage of both information sources to link independent thematic datasets and create rich cartographic applications for data visualization.
This data linking task is generally performed by comparing properties values
of each resource of a given data set, with homologous properties of the resources
described in other datasets [
In order to take advantage of existing data linking tools, we have converted geographic shape data and stored them into a RDF triple store. This task has been achieved by using the Datalift4 platform that also enables to perform the linking process with external published datasets, through the use of Silk5 linking tool. Our linking approach is mainly based on geolocation properties comparison.
Thus we have added to Silk more GIS distance measures for computing the shortest distance between any geometric primitive and simple position locations used in thematic datasets.
The result of this interlinking process is a list of owl:sameAs links between entities of each datasets, at a given threshold. These links are used to extend the geographic reference data set with information queried on the y from the external thematic datasets through the visualization interface. We have applied this approach on a geographical reference dataset about buildings and data about historical monuments extracted from French DBpedia6, on the area of Paris.
The use of links between thematic and reference data could be further investigated to enable data visualization at di erent level of detail, and visual detection errors during matching process of geodata.