Suggestion: make documents optional for BERTopic.transform()

Nobody has claimed this yet.

Hi Maarten,

Thank you once again for your fantastic work :))

I'm using BERTopic as part of a sklearn-style pipeline. Looking at the transform function, it seems that it only uses the documents-input if embeddings are not provided. Therefore it might make sense to make documents an optional parameter. Right not I use a hacky workaround where I provide it with some sham-documents, but this introduces a bit of overhead for large datasets. Would this be possible or would it disturb the API too much?

I would be more than happy to implement it, but it shouldn't be too difficult (if you think it is a good idea)

Start by reading BERTopic's transform function and the path used when embeddings are supplied. Check how documents are currently required in that case and verify the sklearn-style pipeline behavior. Done means callers can omit documents when providing embeddings without the current sham-document overhead, with the existing transform behavior preserved otherwise.

Written by the indexing model from the issue text.