An approach is presented to the automatic discovery of labels of relations be- tween pairs of ontological classes. Using a hyperlinked encyclopaedic resource, we gather evidence for likely predicative la- bels by searching for sentences that de- scribe relations between terms. The terms are instances of the pair of ontological classes under consideration, drawn from a populated knowledge base. Verbs or verb phrases are automatically extracted, yielding a ranked list of candidate rela- tions. Human judges rate the extracted relations. The extracted relations provide a basis for automatic ontology discovery from a non-relational database. The ap- proach is demonstrated on a database from the natural history domain.
Marieke van Erp, Antal van den Bosch, Sander Wubben, Steve Hunt
ILK Research Group
Tilburg centre for Creative Computing