Replies: 1 comment
-
Choosing the number of topics highly depends on the underlying clustering algorithm as each will behave differently. For instance, k-Means with a centroid-based technique handles clustering differently from a density-based algorithm like HDBSCAN. So it would depend on the underlying model that you use. I personally would advise involving human evaluation as much as possible as grid-searching evaluation metrics often leads to misleading results. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Dear Maarten,
I am pretty new to Python and BERTopic. I wonder if BERTopic has any methods that can help users decide the number of topic/clusters, such as perplexity score. Any help is appreciated!
Beta Was this translation helpful? Give feedback.
All reactions