Linking the Thesaurus for the Social Sciences to the Web of Linked Data

Abstract

In this paper, we apply different methods for linking subject headings of the Thesaurus for the Social Sciences (TheSoz) to DBpedia, the nucleus of the Web of Linked Data which is derived from the structured information of Wikipedia. Our method utilizes the backlinks and outlinks within Wikipedia for link detection. We examine to what extent the linking process can be optimized with the help of a network-based similarity measure, in order to achieve a higher precision and recall. We test two baseline methods, string alignment and language property matching and compare them to our own method. Our method outperforms the F-scores of the baselines by 10 percentage points.