Japanese index

["Jan Tosovsky" <j.tosovsky@email.cz>]

Dear All,

has anybody any experience with generating Japanese back-of-the-book index
from DocBook source?

I am facing same issues discussed in this old thread (all entries end up in
the Symbols section):
https://lists.oasis-open.org/archives/docbook-apps/200605/msg00063.html

If I understand correctly, indices in Japanese should be grouped
phonetically:
https://www.slideshare.net/k16shikano/imybp-light

I've found promising Kuromoji library https://github.com/atilika/kuromoji
I can imagine it could somehow pre-process all index entries and generate
values for the 'sortas' attribute. 

But it is still unclear how to tweak the index code to generate groups from
non-latin characters.

Or are there better ways?

Thanks,

Jan