Type inference through the analysis of Wikipedia links.

DBpedia contains millions of untyped entities, either if we consider the native DBpedia ontology, or Yago plus Word-Net. Is it possible to automatically classify those entities? Based on previous work on wikilink invariances, we wondered if wikilinks convey a knowledge rich enough for their classication. In this paper we give three contributions. Concerning the DBpedia link structure, we describe some measurements and notice both problems (e.g. the bias that could be induced by the incomplete ontological coverage of the DBpedia ontology), and potentials existing in current type coverage. Concerning classication, we present two techniques that exploit wikilinks, one based on induction from machine learning techniques, and the other on abduction. Finally, we discuss the limited results of classication, which conrmed our fears expressed in the description of general gures from the measurement. We also suggest some new possible directions to entity classication that could be taken.