Hey forum,
If I have an Author Topic Model (ATM) trained on a set of documents and their authors, and then I have a new sample of documents and the corresponding (different) authors. Can I use an already trained model to infer the distribution for the new set of authors?
I have looked at the function
get_new_author_topics
however, this appears to assume that each new document has only one author. Is that true? That appears to be computing a document level distribution for the cases in which the document is multi-authored.
Thanks!