I am playing around with embedding German texts. The documentation states that the embedding endpoint is working best with English texts.
Is it usable for German at all? I find the similarity score I am getting back quite random. Wondering if I am doing something wrong, or if it doesn’t work with German language anyways.
Unfortunately, embedding is fine tuned for English only (its mentioned in the online docs somewhere, maybe the announcement blog post by OpenAI as well)
It doesn’t use Davinci which has a good understanding of German
ja - I read that its supposed to work best in English. I was just wondering if its like “10% better” or so - but to me it seems right now that it is unusable for German embeddings