How to merge topics for model `BERTopic_Wikipedia`

#2
by benjaminliupenrose - opened

I am using BERTopic_Wikipedia model now and hope to have a customized version without training wikipedia from scratch.

Specifically, I want to use BERTopic topic_model.merge_topics(docs, topics_to_merge) function. However, it required var=docs which is 200GB in size. Is there a workaround or hacky way to merge topics for this specific model?

Currently, this is unfortunately not possible. The documents are needed to create updated topic representations since the topics would change if you merge them, so a different representation would be needed.

Sign up or log in to comment