WIP: Positive/negative highest ranking terms

[x-tabdeveloping]

No description provided.

[rbroc]

looks good! just a couple of comments/questions:

• more of a matter of taste, but for less repeated code the logic that is now in _topics_table could actually be implemented as part of get_topics (as it was before)? not a biggie tho, can leave as is right now and potentially discuss this for later refactoring;
• following the same logic as with keywords, should the user be able to see "lowest" ranking docs for a topic -- should there be an option to return negative documents in _highest_ranking_docs?

[x-tabdeveloping]

Regarding get_topics() I'm not sure about that one. I only included that method so that the uninitiated, who come over from e.g. Gensim, and do not know how topic models work in sklearn can get a nice JSON serializable Python object. So it's basically an interface for people coming from the more software-development side of things who don't like NumPy and Pandas magic.
On the other hand I don't personally like it, and I'm not sure it is worth the hustle to implement negative terms in it. What do you think @rbroc ?

[x-tabdeveloping]

Regarding the other point: Yes, totally! I just haven't implemented it yet. But if you look at the issue's description I already outlined that I wanna do it :D

[rbroc]

Regarding get_topics() I'm not sure about that one. I only included that method so that the uninitiated, who come over from e.g. Gensim, and do not know how topic models work in sklearn can get a nice JSON serializable Python object. So it's basically an interface for people coming from the more software-development side of things who don't like NumPy and Pandas magic.
On the other hand I don't personally like it, and I'm not sure it is worth the hustle to implement negative terms in it. What do you think @rbroc ?

Fine by me to leave as is for now - I like how “semantically” transparent get_topics is but this is something we can re-evaluate later. :)

[rbroc]

LGTM! I’ll leave it to you to decide if for transparency you want to rename show_negative arguments to show_lowest.