Change parameter used in dtm_coherence() (in DTM wrapper) to avoid persistent warning

[MeganStodel]

Within the DTM wrapper, using dtm_coherence() always produces a warning:

"The parameter num_words is deprecated, will be removed in 4.0.0, use topn instead."

Although the function works, obviously the intention is to deprecate the parameter at some point.

This can be tracked back to show_topic(), which is where num_words has switched to topn. It is a very simple fix to just change the parameter used in the dtm_coherence() function accordingly.

PR incoming with fix.

[piskvorky]

Thanks. @MeganStodel what do you use DTM for? What's your motivation, context?

We're seriously considering removing all third party wrappers from Gensim: Mallet, DTM, etc. So any feedback on which parts are still in use & relevant, and why, is appreciated.

[MeganStodel]

Thanks @piskvorky, that's interesting!

I use Mallet regularly and have just started exploring DTM. With DTM, I know there is another implementation in gensim (ldaseqmodel) but I found this much, much slower than using the wrapper. The wrappers in gensim are the only route I know for applying Mallet and DTM models in python, so would definitely feel their loss.

[piskvorky]

Thanks Megan, duly noted.