The method requires GBNC and a thesaurus for a particular language.
iii) a dependency parser and a large text corpus (or the GBNC dataset).
We carried out an additional postprocessing step of the GBNC, a conceptual indexing step.
We also evaluated FE discovery algorithm on English GBNC dataset, using English FrameNet and Wordnet.
For English both GBNC data and and WordNet are available, so we use Wordnet instead of RuThes-lite.