Identify desired plots / layout schemes

[msarahan]

• What variables should be presented in an interface?
• What style plots would be nice?
• What plot styles or tools may need further development?

I'm assuming we'll prefer using Bokeh.

[msarahan]

Cluster visualization (this is LDAVis' strength)

Scaling is critical: number of docs. LDAVis limited in this regard.

Termite plot clusters need to be tied to output of clustering visualization!

LDA: similarity quotient is variable to play with. Highlights significant terms. Nice to have - not in high usage, but cute.

Desire: trending terms over time. Mocked up in Kibana, but Kibana has no route back to raw data. Want ability to go from cluster to raw data that contributed.

Stats to display:

• model stats
• number of docs in corpus
• number of docs containing terms
• number of docs with topic above user-specified threshold for "belonging"

Topic tagging/naming: user specified, but tracked through plotting by name instead of by topic index

[gpfreitas]

For some reason, I can't close this issue.

@msarahan , can you close it? Trying to cleanup my "assigned issues".

[gpfreitas]

Maybe we should just disable the issues in the "Settings" of the repo altogether. I did that with another old project repo, and it worked great (issues were preserved after I re-enabled them as a test... But I'm not sure GH doesn't erase them after a while. I'd be surprised, but if it's important, we should check.)