I don’t think arithmetic coding achieves the 1 bit / character theoretical entropy of common English, as that requires knowledge of very complex boundaries in the probability distribution. If you know a color word is coming next, you can capitalize on it, but not letterwise.
Of course, if you permit a large enough block size, then it could work, but the lookup table would probably be umanageable.
Doesn’t arithmetic coding accomplish this? Or does that not count because it’s unlikely a human could actually use it?
I don’t think arithmetic coding achieves the 1 bit / character theoretical entropy of common English, as that requires knowledge of very complex boundaries in the probability distribution. If you know a color word is coming next, you can capitalize on it, but not letterwise.
Of course, if you permit a large enough block size, then it could work, but the lookup table would probably be umanageable.
Yeah, I meant “arithmetic encoding with absurdly large block size”; I don’t have a practical solution.