Evaluating Topological Queries in Linked Data Using DBpedia and GeoNames in Switzerland and Scotland
Published online on May 20, 2016
Abstract
This article explores the practicability and assesses the quality of searches for Wikipedia pages of topologically related administrative divisions in Switzerland and Scotland via Linked Data. To this end, the quality of searches in the English and German versions of DBpedia is compared, as is that of searches in GeoNames and DBpedia using DBpedia's links to GeoNames or a manually created list of links. In addition, the effects of updates to Swiss municipality borders in the English version of DBpedia on standard quality metrics are studied. It turns out that live searches are practicable with acceptable performance, but that the quality of searches in terms of recall and precision is acceptable only if DBpedia is queried via GeoNames and DBpedia's links to GeoNames are replaced by manually created links. Searches for Scottish unitary authority wards perform poorly for all approaches investigated. Recall and precision are strongly influenced by changes in municipality borders, with a relatively long latency suggesting that, at least in some cases, administrative changes propagate slowly into DBpedia. These results have important implications for those seeking to use location as a means to link data.