Comment by visarga

It would not be inside Wikipedia, but separate, as a collection used in the training corpus.