Lda difference visualization

[menshikh-iv]

You can see result with nbviewer link with interactive graphs

Continuing #1243

[tmylk]

Please explain the annotation with '++' and '--' a little bit more. it's explained already, just need to mention the + and - symbols explicitly.

[menshikh-iv]

@tmylk Done!

[tmylk]

It is better than before but please add a very simple viz where it's just two topics in order to explain the idea more clearly.

[piskvorky]

@menshikh-iv There's no way to review a notebook (on github), so I'll comment here instead:

1. add full stops = . at the end of sentences (many places)
2. prefer list/set/dict comprehensions in place of filter/map
3. use hanging indent
4. use normal numpy array instead of list-of-lists (mdiff)
5. some intro, e.g. who is "I"?
6. missing articles etc -- do we have any students who are native speakers, who could go over this and proofread?

[parulsethi]

@menshikh-iv An example that I think could be helpful for user:

After fitting the LDA models, print an example using show_topics() and then describe that the distance we are calculating is basically between the topic's probability distribution/bag of words that we see above.

Maybe, illustrate the process also by taking an example cell in the matrix - print both topic distribution of that cell and then use distance function on these 2 distribution/words to print the distance which comes out to be of same value as in the matrix cell.