I agree. In my original post I mention that the first step would be to get a list of glyphs, which would probably require manual intervention. (Although not in the simple case where all multi-character glyphs are escaped)

But even if the glyphs are known I can't see how to go forward.

I'm also not necessarily trying for a perfect solution.

btw, I take to heart youre less is more comment. I think the problem could have been much more clearly presented, and I will try to do so in the future.