Thanks to Ralph DeCarli for the link in another thread ( ) to an
open-source machine translation project, Apertium.

Even though machine translation has a long ways to go to become
perfect, as conlangers we are in the unique position of being capable
of having a machine translation application that works perfectly for a

1. Using a temporary grammar create the Apertium files for a language
pair English<->conlang.

2. Discard the temporary grammar documents and declare that henceforth
the corpus produced by the translation ap shall be the final authority
on correct usage.

3. Write a descriptive grammar documenting whatever is produced Apertium.

It follows that every English to conlang translation by Apertium is
correct. It's a simple matter of shooting an arrow at the wall and
then drawing the bullseye around the spot the arrow hits.