Internationally Agreed DTMF to Generic Message Snippet Equivalency Recommendation

When leaving an answering machine message, it'd be useful to be able to compose a wide variety of messages just by pressing the number buttons, interspersed with you saying your name and perhaps the names of various things, places, species, etc.

The answering machine could have dtmf to speech synthesis software (or hardware, if all the cogs fit) and the output language and colloquialisms would suit the localisation settings of the phone. But the semantics would be universally agreed around the world, world wide.

The idea is that each number isn't a complete sentence, but a semi-meaningful component or aggregation of parts of speech enabling sentences to be built up, perhaps using something like the macro equivalent of T9. Thus, at the laziest way, you might default to pressing [and saying] - 1,2,[your name],3,[aardvark],4 and this would give a sentence that covers the most basic and common reason for leaving a message without any significant detail, and yet remain identifiable as a stand-alone entity. More useful combinations would involve the entire base ten range on the traditional upside-down (wrt calculators) keypad, giving a nearly unlimited range of expression. Probably.