Inferring topic distributions on unseen documents with metapy?


#1

Is it possible to infer topic distributions on new, unseen documents with an LDA model in metapy?

I am working on a topic modelling problem and am currently using gensim, however the LDA method used in gensim gives me poor results. So I would like to try using metapy since it has a gibbs sampling implementation of LDA. I cannot find in the documentation or tutorials any obvious way that to infer topic distributions on new, unseen documents.


#2

Yes, this is something that is missing in the current interface in metapy. The corresponding classes just landed in the C++ library, so I will be adding an interface to metapy for them very soon.