How is frequency calculated in topics per class?

[chrislalk]

Hi all!

I wonder how frequency is calculated in topics per class and how does it relate to the probabilities? Intuitively, I would have guessed that I can aggregate the probabilities of a topic over all documents in a class and use this as my expected value for this class similar to frequency. However, then I wouldn't necessarily get integers as in frequency. Not sure though if that makes sense.
Any thoughts?

[MaartenGr]

I wonder how frequency is calculated in topics per class and how does it relate to the probabilities?

The frequencies in the .topics_per_class function is simply the count of documents in that specific class across each topic. In other words, it is the number of documents for topic t in class c. Thus, this frequency has no relation at all with the probabilities.

Intuitively, I would have guessed that I can aggregate the probabilities of a topic over all documents in a class and use this as my expected value for this class similar to frequency.

Could you elaborate a bit more on this? I am not sure if I understood this correctly. Do you want the probability of a topic belonging to a specific class?