How to implement NMF dynamic topic modeling using gensim?

[squaaad yang]

Hi everyone,

Recently, I have been doing a dynamic topic modeling project using NMF. 

I thought the corpus of different time windows was loaded into the model by the "update" command. However, I found that some projects just loaded the whole corpus into the model without splitting them into various slices. I do not know how such a model can differentiate different time windows.

Therefore, I would like to know how I can train a dynamic topic model based on NMF with gensim.

Thanks for any discussion and answers.

YANG