It seems to me that, in order for such an informationally dense language to actually be used, especially as spoken word, it would need to include a layer of error-correcting codes on top of the basic language. That could be a fascinating linguistic concept on its own, especially if one could design the error correcting layer in such a way that a human being (as opposed to a computer) could conceivably become fluent in it.

I would guess the author has thought if this; wonder if it already exists.