Do you think that this is understandable?
https://27139-843222-gh.circle-artifacts.com/0/doc/modules/multiclass.html

It looks reasonable, but a lot of the information is redundant with the glossary section on target types. So perhaps it makes more sense to refer the reader there for details on formats and APIs.

To refer to the glossary, you should use the term role, e.g. :term:`multiclass`

in a conversation at scipy2018, @ivanov suggested that maybe the problem for rewording the multiclass was that the examples were not the 'same' for all the scenarios.

So here's a draft. I'm not really sure that all the examples make sense or are clear enough. So, comments are welcome!!

If anyone can think of a different set of coherent examples, don't be shy.

here's the artifact https://27909-843222-gh.circle-artifacts.com/0/doc/modules/multiclass.html

IMO, I find that the changes make the documentation hard to read (too much level of bullet points)

@lucyleeow can you give it a read? and we finish it

This is the current master https://scikit-learn.org/stable/modules/multiclass.html

here is the last version https://29571-843222-gh.circle-artifacts.com/0/doc/modules/multiclass.html

@massich it's still failing - but this may be because master is broken?

Yes, CI still fails on master cf #14956 (comment) . We need to investigate it.

Closing in favor of #15333 since @lucyleeow did a great job!