@ramon, nope, it’s really interesting. And I gave @ajnyga my ‘like’. He really did a great job, since he even managed to implement a kind of autocomplete based on thesauri used. That’s really a sine qua non for such use cases.
I still have not had time enough to test it, but I think it is an ad hoc solution now, i.e. it is built for usage with that particular thesaurus. So the question is how easy it might be adopted to any other vocabulary, say JEL Classification System (nonSKOS)?