Semantic Relatedness in DBpedia: A Comparative and Experimental Assessment

The evaluation of semantic relatedness between web resources is still an open chal- lenge. This paper focuses on knowledge-based methods that provide an alternative to corpus-based approaches and generally rely on the availability of knowledge graphs. In particular, we selected 10 methods from the existing literature, organized by adjacent resources, triple patterns, and triple weights-based methods. They were im- plemented and evaluated using DBpedia as a reference RDF knowledge graph. Since DBpedia is under constant development, the experimental results of these proposals are not comparable in the literature. For this reason, in this work, these methods were experimented with the same DBpedia release and 14 well-known golden datasets at once. Based on this experimentation, the weighting of RDF triples combined with the evaluation of all the directed paths linking the compared resources is the best strategy for computing semantic relatedness in DBpedia