Dictionary quality is paramount for correction quality. These dictionaries have both reliable frequency values and correct vocabulary. This was achieved by intersecting the large Ngram datasets from google Ngram with wordlists generated from hunspell dictionary files.
Still, it sometimes might be a good idea to generate your own frequency dictionary which better fits your use-case.
The python script used for downloading the Ngram dataset, decompressing it and merging it with existing wordlists is also provided.