Abstract
Natural language definitions of terms can serve as a rich source ofknowledge, but structuring them into a comprehensible semantic model isessential to enable them to be used in semantic interpretation tasks. Wepropose a method and provide a set of tools for automatically building a graphworld knowledge base from natural language definitions. Adopting a conceptualmodel composed of a set of semantic roles for dictionary definitions, wetrained a classifier for automatically labeling definitions, preparing the datato be later converted to a graph representation. WordNetGraph, a knowledgegraph built out of noun and verb WordNet definitions according to thismethodology, was successfully used in an interpretable text entailmentrecognition approach which uses paths in this graph to provide clearjustifications for entailment decisions.