generate german text

#15
by hmrc87 - opened

Hi, is it possible to generate german text with it?

Databricks org

In general it cannot generate text in non-English languages very well. The base model was primarily trained on English text. However, there is still going to be some non-English text in the corpus. For example, there are web pages in English that teach you German expressions. Due to this reason the model may sometimes learn an association between some English and another language. You might find some cases where it will translate some basic phrases okay due to this assocation, but I wouldn't rely on the model for this purpose. I've tested Dolly v1 in these scenarios and it tends to make up sentences that kind of look like German but aren't correct.

Thanks for the quick answer, Matt! I also tested out gpt2 with nanoGPT and ran into the same issues. However it was really funny and entertaining - but definitely not german :D

hmrc87 changed discussion status to closed
Databricks org

Yes I do find the attempts pretty amusing as well :)

Sign up or log in to comment