German translation dataset very formal / in parts incorrect

[HAuser1234]

Describe the bug
The german translation is mostly good, but especially in data/piles/german/pile_of_specific_actions.csv you can find sentences like: „Schließen Sie die Jalousien im Wohnzimmer“. This would very formally beg the assistant to close the blinds, which probably most germans never do. Therefore the sentence should be „Schließe die Jalousien im Wohnzimmer“.
The problem will probably be that most sentences use this extremely formal language.

Also the responses have the same issue, but there ist is partly even more problematic as the responses sound like a command to the user to do a action instead of the assistant doing it for the user :)

example:
data/piles/german/pile_of_responses.csv
Schließen Sie die Jalousien wie gewünscht. -> should be „Ich schließe die Jalousien wie gewünscht“
or some grammatically completely incorrect sentences: Öffnen Sie die Jalousien für Sie. -> should be the same sentence as before.

This I will probably create issues for some users. If you want I probably can correct some of the errors, but it would be great if others could help too.

Edit: using deepL as the translator instead of google works way better, but needs a api key that is limited to 500000 free characters per month!

[acon96]

I don't know German so I can't really comment on any of this. Will explore other translation options (the OpenAI translator pointed at a locally hosted model seemed like a potentially un-throttled way) and will definitely accept manual translations as they are always better.

[HAuser1234]

Yes I tried ChatGPT as a translator and it works (nearly) perfect! When you release the next version that can on my older CPU then I can try the current state and report back. And when you decide to accept manual translations or review other ones let me know!