Automatising the learning of lexical patterns: An application to the enrichment of WordNet by extracting semantic relationships from Wikipedia

less than 1 minute read

by Maria Ruiz-Casado et al.

Ruiz-Casado et al. extend WordNet by extracting hyponym, hyperonym, holonym and meronym relations from the simple english version of WikiPedia. The authors automize

  • the extraction of relations and
  • the process of learning textual patterns modeling these relations (e.g. X is one of the PLURAL-NOUN in Y)
The described approach applies the following steps:

  • word sense disambiguation
  • pattern extraction based on existing relations in WordNet
  • pattern generalization of the patterns retrieved in the step above
  • identification of new relations
The authors show that the precision of the generated patterns is similar to patterns written by hand.