The translation assistant you write is actually very interesting. Heavily rule based but surprising to see it actually helps at all.
But the scale of the project itself seems still pretty limited, reimplementation could still be an option.
Overall good read and interesting approach