Basically, I concur with Alianora and Nathan.
Even if parsers were actually perfect and could understand natural language, it’s not clear that people would want to type out everything in full.
They might, and most probably would, still resort to abbrevations, because it’s more convenient. The only exception might be if everything was voice-driven, because it’s not much of a difference between saying “L” (“ELL”) versus “LOOK” out loud, and it would be more like a real-life conversation. (But it’s not unlikely that even in a voice-input system, abbrevations would get established soon.)
Consider chat (IRC/messenger). When humans are chatting with each other via the net, they use abbreviations a lot, even though their addressee has human-level understanding and would have no problem comprehending even very long and difficult sentences.
Presumably, in the ideal parser world, people would simply use whatever they want. Sometimes, they might write or say: “let me bring the flowers in the vase into disarray in order to annoy my fussy aunt Margret”, and sometimes they might just say “look” or “x table”, and the parser would understand all of it.
But it’s not the conventions which limit the evolution of free-text parsers.
What limits the evolution of parsers is the difficulty of Natural Language Processing and general-purpose AI.
Edit: Partly ninja’d by Natrium729.