ABSTRACT
With the ongoing initiatives aimed at publishing data in a linked data format [1] (e.g., DBpedia, Freebase, and Wikidata), more and more graph-based representations of knowledge have become available. The graph-based representation turns plain strings into “entities” that possess attributes, taxonomy, and relationships to other entities. Viewed this way, linked data can introduce new capabilities and opportunities for exploring data and revealing interesting connections among entities. However, the enormous volumes of such knowledge bases raise new challenges as well, and they require an inevitable need for embracing semisupervised or unsupervised techniques.
