I haven’t used accents/collation before but if you haven’t configured it to store full documents in the index, you would be getting your actual documents after the search. and then you could work on the result set as usual without fancy additions. Even if you stored them, I think they are stored as is, only the keywords for the index should be changing for the accents.
Can you give us some accented documents to work with so we may at least try to see it from your viewpoint?