Since reading doesn’t actually involve a letter-by-letter examination of the material, I wonder how a lossy compression algorithm that used words with the righ

Since reading doesn’t actually involve a letter-by-letter examination of the material, I wonder how a lossy compression algorithm that used words with the right shape would do compared to lossless compression?