It seems they have pre-processed images of etymologies of most of the words and just indexing it.