Let me answer your last comment directly here, it would be too long to fit the comment section. Your question seems a bit opinion based so my answer can call some discussions afterward if you do not agree with it.
It seems to me that you base the security of your idea on the following concept: two individuals want to communicate, and they share a set of common language that they both understand. To do that, the first player transforms his message $m$ into a message $m'$ which is $m$ cut in carefully chosen blocks, each translated to one of the common languages of the players. Then, $m'$ is encrypted and authenticated as usual.
My claim is that this $m \rightarrow m'$ step does not really add any security to the protocol. To see that, imagine an attacker who has already compromised, in one way or another, the encrypt and authenticate part; meaning, the attacker has recovered $m'$.
There are two possibilities. First, suppose the attacker knowns the set of languages used by the players. If so, he can easily build a software with full access to dictionaries and translators for all of those languages. Given that, he has essentially as much chance as the second player (the player that must receive $m$), as he will follow the same strategy than him to recover $m'$: scan from your knowledge of each language until you find a valid interpretation for the meaning of a block, then mode to the next block. With a software, this should really be trivial. Security is not about adding, perhaps a few hours of work, but about making an attack essentially infeasible given any reasonable time and computational power.
Suppose now that the attacker does not know the common set of languages of the players, and so they can use this as follows: use words that have meanings in several languages, so that their interpretation will be ambiguous for the attacker which does not know the set of languages. First, it should probably be still easy to attack with the same program, identifying all the valid languages and interpretations in those languages, and listing all the possible outcome. Second,the main observation is that using this shared secret information to ensure secrecy is simply equivalent to sharing a common password, that both players know by heart. But it is trivial to observe that one can always encrypt a message with a simple password that is only stored in his head before sending it, to ensure that "some" level of security is maintained. And if this is what you want to do, then the set of common languages is a very bad choice of password: browsing the net, doing some investigations, will probably make it very easy for an attacker to have a good knowledge of which languages the two players are likely to speak.
All in all, it boils down to this: in cryptography, you want systems that provably defeat all attacks, even mounted with powerful software and lots of computational power. Using such simple, human based scrambling of a message is very unlikely to add any concrete security to your system.