In the end I went over all docs one-by-one and fixed stuff manually.
I understand but I find it strange that even if I would give up on processing performance, I can't do a simple text search on documents. (Given the analyzers are not set up accordingly to begin with.)
Anyway, thanks @lukas. I'm looking forward for the feature to be able to copy the queries in all kinds of visualizations. 
That would make our lives so much easier. Btw, is there a github issue for this? I'd give it a thunbs up.