You already have something similar with the current system. Languages like Chinese have more semantic load per character, so I'd imagine Chinese tweeters can tweet more complete thoughts.

A better (but more opaque) system might try to estimate the "quantity" of semantic content in a language-specific way.