The German backtranslator, used in NVDA, for example, has a lot of errors where characters or character combinations are expanded from their contractions inappropriately. That is, there are words in German where there are character combinations that are not appropriate to expand, andnot meant to be backtranslated. A few examples:

• konnt ⠅⠞
  • Appropriate: konnte ⠅⠞⠑, einkonnt ⠫⠅⠞
  • Not appropriate: projekt ⠟⠚⠑⠅⠞ (backtranslated as projekonnt), kontakt ⠅⠕⠝⠞⠁⠅⠞ (backtranslated as kontakonnt)
• falls ⠋
  • appropriate: ebenfalls ⠑⠃⠉⠋, gleichfalls ⠛⠹⠋
  • Not appropriate: einfuhr ⠫⠋⠥⠓⠗ (backtranslated as einfallsuhr), hinfort ⠓⠔⠋⠢⠞ (backtranslated hattinfallsort - see below)
• hatt ⠓
  • Only appropriate in hatte ⠓⠑, hattest ⠓⠑⠾, hattet ⠓⠑⠞, but not the name Hattingen ⠓⠁⠞⠞⠔⠛⠉. This is backtranslated correctly as there is a specific rule for it in the German table.
  • Not appropriate in any other case where ⠓ is at the beginning of a word. See hinfort ⠓⠔⠋⠢⠞, which is being backtranslated as hattinfallsort (two backtranslation errors in one word).

This goes on and on and on for many of the abbreviations which are of one or two forms or in the middle of words and should be treated literally instead of being expanded. ⠥ for ung is a particularly bad example as well.

So, two things should happen:

• If there isn't, a way must be devised to specify that, if all exceptions have been dealt with, the letter should be treated literally, not translated back.
• German should be added to the tests to make sure translation and backtranslation are covered.

I see that the German tables were last touched when initially added in 2009. I can help with this, but it would be good if some other German-speaking and brailling community members could also help with this.